A Wittgenstein Dictionary

Hans-Johann Glock

Editor
Notes on the use of this book

This book addresses three kinds of readers. Academics working inside or outside philosophy
should find explanations of key terms and issues in Wittgenstein's work, and be able to find out
what impact it might have on their own. At the end of entries, I sometimes indicate briefly what
impact it has actually had, but for detailed information one should consult the items listed in the
Bibliography of Secondary Sources. Students working on Wittgenstein or related topics (Frege,
Russell, philosophical logic, metaphysics, epistemology, philosophy of mind) should find an
account of major exegetical and substantive problems. Wittgenstein scholars should find a state-
of-the-art discussion, as well as some n e w ideas. I have striven for a comprehensive coverage of
the topics, but it is synthetic a priori that I have failed.

Readers without prior knowledge of analytic philosophy are strongly advised to read the
Sketch of an Intellectual Biography, by w a y of an introduction. I have tried to avoid formal
symbolism, but some of it is inevitable in discussing logical and mathematical issues. The entries
concerned are easier to understand with some knowledge of the propositional and predicate
calculus. In line with Wittgenstein, I employ the notation of Principia Mathematical, although I
use parentheses instead of dots as scope-indicating devices, and '(3x)(3y)xRy instead of
'(3xy)xRy. Other technical devices are explained in the entries concerned.

Inevitably, many of the interpretations presented here are controversial. I have tried to state
alternative readings which are plausible or widespread. In the end, readers will have to decide for
themselves by looking at Wittgenstein's texts. To facilitate further study, I have not only quoted
famous passages, but have also provided ample references, including references to the Nachlass
where it provides significant additional material. I have m a d e liberal use of cross-references,
which are indicated by the relevant entry title (or a cognate — e.g., 'determinate'/'determinary of
sense', 'grammatical p r o p o s i t i o n ' / ' g r a m m a r ' ) in small capitals. Terms which do not occur as
independent entries can be located from the index.

Unlike other, more distant philosophers in this series, Wittgenstein provides a 'live option'
for contemporary philosophers. For this reason I have striven to convey, wherever possible, h o w
his remarks might be attacked or defended. Some commentators suggest that Wittgenstein does
not engage in a rational debate with other philosophers, but merely tries to convert them to his
point of view. These commentators find his w o r k so out of the ordinary as to be
incommensurable with the rest of philosophy. In m y view this interpretation is unfounded.
Although Wittgenstein's philosophical method is revolutionary in seeking to undermine even the
assumptions underlying previous debates, he does so by w a y of arguments which can be assessed
for their soundness. I have therefore stressed not only lines of historical influence, but
agreements and disagreements with past and present thinkers.

Another view is that Wittgenstein's remarks often do not present answers to his self-posed
questions, or hard-line positions, that they are full of qualifications and investigate rather than
affirm or deny. There is some truth in this view. However, since this is a w o r k of reference I
have tried to present as clear-cut a position as Wittgenstein's prudent qualifications allow.
Perhaps some of the views presented here should die the death of a thousand qualifications, and
others the less-protracted death of straightforward refutation. The task of the continuing debate
about the nature and merit of Wittgenstein's philosophy is to deliver or parry such blows; and the
purpose of this Dictionary is to facilitate that debate.
NOTES ON THE USE OF THIS BOOK

defended. Some commentators suggest that Wittgenstein does not engage in

a rational debate with other philosophers, but merely tries to convert them
to his point of view. These commentators find his work so out of the ordin­
ary as to be incommensurable with the rest of philosophy. In my view this
interpretation is unfounded. Although Wittgenstein's philosophical method is System of reference and primary
revolutionary in seeking to undermine even the assumptions underlying pre­
vious debates, he does so by way of arguments which can be assessed for sources
their soundness. I have therefore stressed not only lines of historical influ­
ence, but agreements and disagreements with past and present thinkers.
Another view is that Wittgenstein's remarks often do not present answers
to his self-posed questions, or hard-line positions, that they are full of qualifi­
cations and investigate rather than affirm or deny. There is some truth in
this view. However, since this is a work of reference I have tried to present Unless otherwise indicated, all references are to pages of the edition cited. I
as clear-cut a position as Wittgenstein's prudent qualifications allow. Perhaps refer to works by Wittgenstein (including JVachlass, lectures, correspondence,
some of the views presented here should die the death of a thousand qualifi­ dictations and works by Waismann derived from Wittgenstein) by the famil­
cations, and others the less-protracted death of straightforward refutation. iar capital-letter system; to works of authors that influenced him by abbre­
The task of the continuing debate about the nature and merit of Wittgen­ viated titles. I have provided my own translations wherever appropriate.
stein's philosophy is to deliver or parry such blows; and the purpose of this References to the giants of yore follow established systems. References to
Dictionary is to facilitate that debate. Kant, for example, are to pages of the first (A) or second (B) edition of the
Critique of Pure Reason.

Wittgenstein's works

1. Articles and books in order of composition

The date of composition is specified in square brackets where appropriate.
RCL 'Review of Coffey, The Science of Logic', The Cambridge Review, 34
(1913) 351; reprinted in PO.
NL 'Notes on Logic' [1913], in NB 93-107.
NM 'Notes dictated to G. E. Moore in Norway' [1914], in NB 108-19.
NB Notebooks 1914-16 [German-English parallel text], ed. G. E. M.
Anscombe and G. H. von Wright, tr. G. E. M. Anscombe, rev.
edn (Oxford: Blackwell, 1979).
Tagebiicher 1914-16 (Frankfurt: Suhrkamp, 1984).
GT Geheime Tagebiicher, ed. W. Baum (Vienna: Turia & Kant, 1991).
These contain remarks from the Notebooks written in a secret
code which have been omitted from NB, and are mainly of
biographical relevance.
PT Proto-Tractatus [1917, German-English parallel text], ed. B. F.
McGuinness, T. Nyberg and G. H. von Wright, tr. D. F. Pears
and B. F. McGuinness, with an introduction by G. H. von
Wright (London: Routledge & Kegan Paul, 1971).

2 3
 SYSTEM OF REFERENCE AND PRIMARY SOURCES
TLP Tractatus Logko-Phiksophicus [German-English parallel text], tr. D.
F. Pears and B. F. McGuinness (London: Roudedge & Kegan
Paul, 1961). References are to numbered sections.
Tractatus Logico-Phihsophicus [German-English parallel text], tr. C.
K. Ogden and F. P. Ramsey (London: Roudedge, 1990). First
published 1922.
Logisch-Phiksophische Abhandlung, Kritische Edition, ed. B. McGuin­
ness and J. Schulte (Frankfurt: Suhrkamp, 1989). First German
edition in Annalen der Naturphiksophie, 14 (1921), 185-262.
WV Worterbuch fur Volksschulen (Vienna: Holder-Pichler-Tempsky, 1926);
facsimile reproduction with an introduction by A. Hiibner 1977.
RLF 'Some Remarks on Logical Form', Proceedings of the Aristotelian So­
ciety, suppl. vol. DC (1929), 162-71.
CV Culture and Value [German-English parallel text], ed. G. H. von
Wright in collaboration with H. Nyman, tr. P. Winch (Oxford:
Blackwell, 1980).
Vermischte Bemerkungen (Frankfurt: Suhrkamp, 1984).
PR Philosophical Remarks [1929-30], ed. R. Rhees, tr. R. Hargreaves
and R. White (Oxford: Blackwell, 1975).
Philosophische Bemerkungen (Frankfurt: Suhrkamp, 1984).
PG Philosophical Grammar, ed. R. Rhees, tr. A. J. P. Kenny (Oxford:
Blackwell, 1974).
Phiksophische Grammatik (Frankfurt: Suhrkamp, 1984).
GB 'Remarks on Frazer's "The Golden Bough"', ed. R. Rhees, Synth-
ese, 17 (1967), 233-53; references are to the complete version in
PO.
BB The Blue and Brown Books [1933-35] (Oxford: Blackwell, 1958).
EPB Eine Philosophische Betrachtung [1936], ed. R. Rhees, Schrifien 5
(Frankfurt: Suhrkamp, 1970), 117-237.
CE 'Cause and Effect: Intuitive Awareness', ed. R. Rhees, tr. P.
Winch, Phiksophia, 6 (1976), 392-145; reprinted in PO.
RFM Remarks on the Foundations of Mathematics [1937-44], ed. G. H. von
Wright, R. Rhees and G. E. M. Anscombe, tr. G. E. M. An­
scombe, rev. edn (Oxford: Blackwell, 1978; 1st edn 1967).
Bemerkungen zu den Grundlagen der Mathematik (Frankfurt: Suhrkamp,
1984).
PI Phiksophkal Investigations [German-English parallel text], ed. G. E.
M. Anscombe and R. Rhees, tr. G. E. M. Anscombe (Oxford:
Blackwell, 1958; 1st edn 1953). References are to sections of
Part I (except for footnotes), and to pages of Part II.
RPP I Remarks on the Philosophy of Psychokgy [1945-7, German-English par­
allel text], volume I, ed. G. E. M. Anscombe and G. H. von
Wright, tr. G. E. M. Anscombe (Oxford: Blackwell, 1980).
4 5
REFKRE N C E AND PRIMARY SOURCES
RPP II Remarks on the Phiksophy of Psychokgy [1948, German-English paral­
lel text], volume II, ed. G. H. von Wright and H. Nyman, tr.
C. G. Luckhardt and M. A. E. Aue (Oxford: Blackwell, 1980).
Bemerkungen zur Phiksophie der Psychokgie (Frankfurt: Suhrkamp,
1984).
Z Zettel [1945-8, German-English parallel text], ed. G. E. M. An­
scombe and G. H. von Wright, tr. G. E. M. Anscombe (Oxford:
Blackwell, 1967).
LW I Last Writings on the Phiksophy of Psychokgy [1948-9, German-English
parallel text], volume 1, ed. G. H. von Wright and H. Nyman,
tr. C. G. Luckhardt and M. A. E. Aue (Oxford: Blackwell,
1982).
Letzte Schrifien zur Phiksophie der Psychokgie (Frankfurt: Suhrkamp,
1984).
LW II Last Writings on the Phiksophy of Psychokgy [1949-51, German-Eng­
lish parallel text], volume 2, ed. G. H. von Wright and H. Ny­
man, tr. C. G. Luckhardt and M. A. E. Aue (Oxford: Blackwell,
1992).
OC On Certainty [1951, German-English parallel text], ed. G. E. M.
Anscombe and G. H. von Wright, tr. D. Paul and G. E. M.
Anscombe (Oxford: Blackwell, 1969).
Uber Gewifiheit (Frankfurt: Suhrkamp, 1984).
ROC Remarks on Colour [1951, German-English parallel text], ed. G. E.
M. Anscombe, tr. L. L. McAlister and Margarete Schattle (Ox­
ford: Blackwell, 1980; 1st edn 1977).
Bemerkungen uber die Farben (Frankfurt: Suhrkamp, 1984).
PO Philosophical Occasions [German-English parallel texts where appro­
priate], ed. J. Klagge and A. Nordmann (Indianapolis: Hackett,
1993). Contains reprinted versions of RCL, RLF, LE, M, LSD,
LPE, CE, LFW, NPL. Unless otherwise specified, these are
cited after the original paginations, which are given in this
anthology.
2. Lectures and conversations
WVC Ludwig Wittgenstein and the Vienna Circle [1929-32], shorthand notes
recorded by F. Waismann, ed. B. F. McGuinness (Oxford:
Blackwell, 1979).
Ludwig Wittgenstein und der Wiener Kreis (Oxford: Blackwell, 1967 and
Frankfurt: Suhrkamp, 1984).
LE 'A Lecture on Ethics' [1929], Phiksophkal Review, 74 (1965), 3-12.
M Wittgenstein's Lectures in 1930-33', in G. E. Moore, Phiksophkal
 SYSTEM OF REFERENCE AND PRIMARY SOURCES SYSTEM OF REFERENCE AND PRIMARY SOURCES

Papers (London: Allen and Unwin, 1959), references are to the 4. Works derived from dictations by or conversations with
reprinted version in PO. Wittgenstein
LWL Wittgenstein's Lectures, Cambridge 1930-1932, from the notes of J.
PLP F. Waismann, The Principks of Linguistic Phiksophy, ed. R. Harre
King and D. Lee, ed. Desmond Lee (Oxford: Blackwell, 1980).
(London: Macmillan, 1965).
AWL Wittgenstein's Lectures, Cambridge 1932-1935, from the notes of A.
LSP F. Waismann, Logik, Sprache, Phiksophu, ed. G. P. Baker and B. F.
Ambrose and M. MacDonald, ed. A. Ambrose (Oxford: Black-
well, 1979). McGuinness (Stuttgart: Reclam, 1976).
FW Dictations k F, Waismann, ed. G. P. Baker (London: Roudedge,
LSD "The Language of Sense Data and Private Experience - Notes
forthcoming). References to dictation numbers.
taken by R. Rhees of Wittgenstein's Lectures, 1936', Philosophical
WAM N. Malcolm, Ludwig Wittgenstein - A Memoir, 2nd edn (Oxford: Ox­
Investigations, 7 (1984), 1-45, 101-40.
ford University Press, 1984; 1st edn 1958).
LPE 'Wittgenstein's Notes for Lectures on "Private Experience" and SDE R. Rhees, 'Some Developments in Wittgenstein's View of Ethics',
"Sense Data'" [1936], ed. R. Rhees, Phiksophkal Review, 77
Phiksophkal Review, 74 (1965), 17-26.
(1968), 275-320.
RR R. Rhees, 'On Continuity: Wittgenstein's Ideas 1938', in R. Rhees,
LC Lectures and Conversations on Aesthetics, Psychokgy and Religious Belief
Discussions of Wittgenstein (London: Routledge & Kegan Paul,
[1938-46], ed. C. Barrett (Oxford: Blackwell, 1966).
LFM Wittgenstein's Lectures on the Foundations of Mathematics, Cambridge 1939, 1970), 104-57.
from the notes of R. G. Bosanquet, N. Malcolm, R. Rhees and RW R. Rhees (ed.), Recolkctions of Wittgenstein (Oxford: Oxford Univer­
Y. Smythies, ed. C. Diamond (Hassocks: Harvester Press, 1976). sity Press, 1984).
LFW 'Lectures on Freedom of the Will' [1939], from the notes of Y. WC O. K. Bouwsma, Wittgenstein: Conversations 1949-1951, ed. J. L.
Smythies, PO 427-44. Craft and R. E. Hustwit (Indianapolis: Hackett, 1986).
NPL 'Notes for the Philosophical Lecture' [1941], ed. D. Stern, PO
445-58. 5. Correspondence
LPP Wittgenstein's Lectures on Phiksophkal Psychokgy 1946-47, notes by P.
T. Geach, K. J. Shah and A. C. Jackson, ed. P. T. Geach EL Letters to Engelmann.
(Hassocks: Harvester Press, 1988). FL Letters to von Ficker.
ML Letters to Moore.
OL Letters to Ogden.
3. Anthologies and collections RAL Letters to Ramsey.
RUL Letters to Russell.
Schrifien (Frankfurt: Suhrkamp):
Vol. 1 (1960): TLP, NB, PI. Vol. 2 (1964): PR. Vol. 3 (1967): WVC. Vol. 4 These letters are quoted by date as specifically as possible. They are col­
(1969): PG. Vol. 5 (1970): BB, EPB, Z. Vol. 6 (1973): RFM. Vol. 7 (1978):
lected in the following editions:
LFM. Vol. 8 (1982): RPP I, RPP II.
Briefe, ed. B. F. McGuinness and G. H. von Wright. Correspondence with
Werkausgabe (Frankfurt: Suhrkamp, 1984):
B. Russell, G. E. Moore, J. M. Keynes, F. P. Ramsey, W. Eccles, P. En­
Vol. 1: NB, TLP, PI. Vol. 2: PR. Vol. 3: WVC. Vol. 4: PG. Vol. 5: BB, gelmann and L. von Ficker. In German, with original version of Wittgen­
EPB. Vol. 6: RFM. Vol. 7: RPP I, RPP II, LW I. Vol. 8: ROC, OC, Z,
CV. stein's own letters (when in English) in an appendix; German translations
J. Schulte (Frankfurt: Suhrkamp, 1980).
The Wittgenstein Reader, ed. A.J. P. Kenny (Oxford: Blackwell, 1994): Letters to C. K. Ogden, ed. G. H. von Wright, with an appendix containing let­
Selections from TLP, BT, PG, BB, LPE, LC, PI, RFM, RPP I & II, Z, OC. ters by F. P. Ramsey, 1923-4 (Oxford: Blackwell/London: Routledge,
1973).
Letters from Ludwig Wittgenstein, with a Memoir by Paul Engelmann, ed. B. F.
McGuinness, tr. L. Furtmuller (Oxford: Blackwell, 1967).
Brufe an Ludwig von Fkker, ed. G. H. von Wright and W. Methlagl (Salzburg:

6 7
 SYSTEM OF REFERENCE AND PRIMARY SOURCES SYSTEM OF REFERENCE AND PRIMARY SOURCES

Muller, 1969); Eng. trans., 'Letters to Ludwig von Ficker', ed. Allan Janik, 'Function' 'Function and Concept'.
tr. B. Gillette, in Wittgenstein: Sources and Perspectives, ed. C. G. Luckhardt 'Sense' 'Sense and Reference'.
(Hassocks: Harvester Press, 1979). 'Concept' 'On Concept and Object'.
Letters to Russell, Keynes and Moore, ed. G. H. von Wright, Eng. trans. B. F. 'Negation' 'Negation'.
McGuinness (Oxford: Blackwell, 1974). 'Thought' "The Thought'.
'Compound' 'Compound Thought'.
All of these papers are in Collected Papers, ed. B. McGuin­
6. Nachlass
ness (Oxford: Blackwell, 1984). They are cited after the
All references to unpublished material follow von Wright's catalogue (G. H. original paginations, which are given in this collection.
von Wright, Wittgenstein (Oxford: Blackwell, 1982), 35ff.). They are by MS or Posthumous Posthumous Writings, ed. H. Hermes, F. Kambartel and F.
TS number followed by page number. I use the following abbreviation: Kaulbach, tr. P. Long and R. White (Oxford: Blackwell,
BT The 'Big Typescript' (TS 213), partly in PO 160-99. 1979).
CorrespondencePhilosophical and Mathematical Correspondence, ed. B. McGuin­
The Nachlass is kept in the Library of Trinity College, Cambridge. It is ness, tr. H. Kaal (Oxford: Blackwell, 1980).
available on microfilm/photocopies from Cornell University, the so-called
'Cornell copy'. The full Nachlass will be available on CD-ROM from Oxford Hertz
University Press, edited by the Wittgenstein Archives at the University of The Principles of Mechanics, tr. D. E. Jones and J. T. Walley
Mechanics
Bergen. The early parts of the Nachlass are currendy being published as the (London: Macmillan, 1899).
Wiener AusgabeI Vienna Edition, ed. M. Nedo (Vienna/New York: Springer, Die Prinzipien der Mechanik (Leipzig: Barth, 1894).
1994-). This edition contains the original pagination. In addition to an
introduction and concordance volumes, it will comprise the following: vol. 1: James
MSS 105, 106; vol. 2: MSS 107, 108; vol. 3: MSS 109, 110; vol. 4: MSS
111, 112; vol. 5: MSS 113, 114; vol. 6: TSS 208, 210; vol. 7/1-2: TS 211; Psychology The Principles of Psychology (New York: Dover, 1950; 1st edn
vol. 8: TS 209 (PR); vol. 9/1-2: TS 212; vol. 10/1-2: TS 213 (BT); vol. 11: 1890).
MSS 153a-b, 154, 155.
Kohler
Gestalt Gestalt Psychology (New York: Mentor, 1975; 1st edn 1930).
Works o f other authors
Mauthner
Boltzmann Beitrdge
Physics Theoretical Physics and Philosophical Problems, ed. B. McGuinness Beitrdge zu einer Kritik der Sprache (Stuttgart: Cotta, 1901-3).
(Dordrecht: Reidel, 1974).
Writings Moore
Frege
Notation Selected Writings, ed. T. Baldwin (London: Roudedge, 1994).
Conceptual Notation and Related Articles, tr. and ed. T. W. By-
num (Oxford: Clarendon, 1972). Mathematics Ramsey
Begnffsschrift (Halle: Nebert, 1879).
Foundations The Foundations of Mathematics and Other Logical Essays (London:
The Foundations of Arithmetic, or. J. L. Austin, 2nd edn (Oxford:
Blackwell, 1953; 1st edn 1950). Routledge & Kegan Paul, 1931).
Russell
Die Grundlagen der Arithmetik (Breslau: Koebner, 1884). Principles
Laws The Principles of Mathematics, 2nd edn (London: Allen and Un-
The Basic Laws of Arithmetic, tr. and ed. M. Furth (Berkeley/ win, 1937; 1st edn 1903).
Los Angeles: University of California Press, 1964). Essays Philosophical Essays, 2nd edn (London: Roudedge, 1994; 1st
Grundgesetze der Arithmetik (Jena: Pohle, 1893 and 1903). edn 1910).

8 9
 SYSTEM OF REFERENCE AND PRIMARY SOURCES

Princtpia Principia Matkematica, 2nd edn (Cambridge: Cambridge Uni­

versity Press, 1927; 1st edn 1910).
Problems The Problems of Philosophy (Oxford: Oxford University Press,
1980; 1st edn 1912).
'Theory' 'The Theory of Knowledge' [1913], in The Collected Papers of Sketch of an intellectual biography
Bertrand Russell, vol. 7, ed. E. Eames and K. Blackwell
(London: Allen and Unwin, 1984).
External Our Knowledge of the External World as a Field for Scientific Method
in Philosophy, rev. edn (London: Roudedge, 1993; 1st edn
1914). Ludwig Wittgenstein (1889-1951) was the youngest child of a wealthy and
Mysticism Mysticism and Logic (London: Longmans, Green, 1918). cultured Viennese family of Jewish descent. The Wittgenstein home was a
Introduction Introduction to Mathematical Philosophy (London: Allen and Un­ cenore of artistic and, in particular, musical life. It provided Ludwig with
win, 1919). what he later called his 'good intellectual nursery-training', which consisted
'Introduction' 'Introduction' to Tractatus Logico-Phihsophicus (TLP). of the music of Viennese classicism and a strand of German literature -
Analysis The Analysis of Mind (London: Allen and Unwin, 1921). with Goethe as a figurehead - which rejected the nationalism and faith in
'Limits' 'The Limits of Empiricism', Proceedings of the Aristotelian Society, progress that characterized the mainstream of European culture in the nine­
XXXVI (1935-6). teenth and early twentieth centuries. Wittgenstein was a cultural con­
Logic Logic and Knowledge, ed. R. C. Marsh (London: Allen and Un­ servative who felt at odds with the 'spirit of the main current of European
win, 1956). and American civilization' (CV 6-7; CV contains Wittgenstein's intermittent
reflections on cultural questions). But his intense intellectual passion and
Schlick
honesty prevented him from being nostalgic or parochial. Indeed, he reacted
Papers Philosophical Papers, ed. H. L. Mulder and B. F. B. van der in a highly creative way to certain modern ideas. This becomes clear when
Velde-Schlick (London: Reidel, 1979). we turn to the direct influences on his thinking, which he listed in 1931:
Boltzmann, Hertz, Schopenhauer, Frege, Russell, Kraus, Loos, Weininger,
Schopenhauer
Spengler, Sraffa (CV 19). Those which are relevant to his earlier philosophy
World The World as Will and Representation, tr. E. F. J. Payne (New fall into three groups: the sages, the philosopher-scientists and the philoso­
York: Dover, 1966; 1st edn of trans. 1958). pher-logicians.
Die Welt als Wille und Vorstellung (Leipzig: Brockhaus, 1844; 1st
edn 1819).
Sages, Scientists and. M a d m e n
The sages were thinkers outside academic philosophy whose work Wittgen­
stein read as a youngster. Karl Kraus, the formidable cultural critic of the
late Habsburg Empire, impressed Ludwig by his insistence on personal
integrity. Wittgenstein was also influenced by Kraus's masterful polemical
analysis of language. Opponents are literally taken at their word. Their style,
sometimes even a single ill-judged sentence, is taken to reveal both their fal­
lacies and their character failings. Kraus's work was part of the so-called
'crisis of language', a general concern with the authenticity of symbolic
expression in art and public life. Another expression of this crisis was
Mauthner's critique of language. Mauthner pursued a Kantian goal, the
defeat of metaphysical speculation. But he surplanted the critique of reason
with a critique of language, and his work owed more to Hume and Mach.
His method was psychologistic and historicist: the critique of language is

10 11
 A
aesthetics Aesthetics did not lie at the centre of Wittgenstein's philosophi­
cal interests; but art, especially music, had a paramount place in his life.
While his taste in music and literature tended to be conservative, the house
he designed for his sister Margarete in Vienna in 1926 was modernist. Its
extreme austerity radicalizes the anti-decorative ideal of the Austrian archi­
tect Adolf Loos, whom Wittgenstein at one time admired. By far Wittgen­
stein's greatest contribution to art, however, is his writing, which is one of the
few highlights of German philosophical prose, albeit an exotic one. He had
self-professed aesthetic ambitions, and regarded 'correct' style as integral to
good philosophizing (CV 39, 87; Z §713). His writing is not discursive, but
consists of short and often laconic remarks. Wittgenstein's similes and analo­
gies, and his elusive wit, are reminiscent of Lichtenberg. However, his
remarks are not isolated aperqus, but part of a philosophical line of thought. In
Tractates Logko-Phiksophicus they are very dense, and fitted into a complex
structure of great architectonic appeal, while Phiksophkal Investigations is more
colloquial.
In spite of his personal interest, Wittgenstein's early remarks on aesthetics
are cryptic applications of a philosophical system, his version of Scho­
penhauer's transcendental idealism. 'Ethics and aesthetics are one' (TLP
6.421). This sibylline pronouncement involves three points. Firsdy, like logic
and ethics, aesthetics is concerned not with contingent matters of fact, but
with what could not be otherwise. Hence it cannot be expressed in mean­
ingful (BIPOLAR) propositions, but only shown (NB 24.7.16; TLP 6.13). Sec­
ondly, together with ETHICS, aesthetics constitutes the 'higher' realm of values.
It is transcendent, since values 'cannot lie within the world', but are located
in a Schopenhauerian metaphysical WILL outside it (TLP 6.41-6.432; NB
2.8.16). Finally, like logic, ethics and aesthetics are based on a MYSTICAL
experience, namely marvelling not at how the world is, but that it is. In doing
so, I view the world from the outside, as a 'limited whole'. In addition,
ethics and aesthetics involve 'looking at the world with a happy eye', that is,
with a Stoic acceptance of facts which are not subject to the will. The 'work
of art' is 'the object seen sub specie aeternitatis'. This is reminiscent of Scho­
penhauer's idea that in aesthetic contemplation we escape the domination of

31
AESTHETICS
the will (of our desires), since our consciousness is filled by a single image. It
also links the aesthetic perspective to SOLIPSISM: in viewing the object, or the
world, sub specie aetemitatis I make it my own (NB 19.9./7.10./8.10./
20.10.16; TLP 5.552, 6.43, 6.45; cf. World I §34).
Wittgenstein's early remarks on aesthetics are important to his mysticism,
but shed litde light on art. The identification of ethics and aesthetics under
the umbrella of ineffable values obscures precisely the kind of conceptual
differences he later tried to emphasize. Arguably, neither Schopenhauer nor
the early Wittgenstein adds much beyond metaphysical mystery-mongering
to Kant's insight that aesthetic appreciation involves 'disinterested con­
templation'. Wittgenstein's later discussion yields more palpable results. He
abandons the idea that aesthetic value is ineffable, observes that 'the subject
(Aesthetics) is very big and entirely misunderstood' (LC 1), and points to
four major mistakes:
(a) It is wrong to focus just on a small group of terms like 'beautiful' or
'ugly'. These are used mainly as interjections, and have 'almost a negligible
place' in our reaction to works of art or natural beauty. Most of our aes­
thetic appreciation consists not in simply liking or disliking a work of art,
but in understanding or characterizing it. And where we do assess a work of
art, it is not so much as beautiful or hideous, as as right or wrong, closer or
more distant to certain ideals or standards. Finally, there are 'tremendous'
masterpieces, such as Beethoven's symphonies, which set their own stan­
dards, and impress us almost in the way spectacular natural phenomena do
(LC 1-11; CV 54-5).
(b) It is wrong to neglect the USE of aesthetic expressions in favour of their
linguistic form. Aesthetic appreciation evolves from reactions like delight or
discomfort. What matters is not so much words as the occasions on which
they are used. These occasions in turn are part of an 'enormously compli­
cated situation'. They must be seen against the background of certain activ­
ities, and ultimately of a certain culture or even FORM OF LIFE (LC 2). In
describing musical taste, for example, one may have to describe the social
role played by musical performances. Unfortunately, Wittgenstein did not
clarify whether this context consists exclusively of the social role of the arte­
fact, or whether it also includes, for example, the intentions of the artist.
What is clear, as in the case of ethics, is that his CONTEXTUALISM has relativis-
tic implications. Although Wittgenstein speaks of 'deterioration' within an
artistic tradition, such as German music, aesthetic standards cannot be
judged externally. One may not even understand how to appreciate the
works of an artistic tradition (e.g., African art) without immersing oneself in
the relevant culture (LC 1-11; LW I §§750-3; PI II 230; CE 399).
32
AESTHETICS
(c) The most straightforward aspect of Wittgenstein's aesthetics is his appli­
cation of the idea of FAMILY RESEMBLANCE. He rejects the craving for an ana­
lytic definition of aesthetic terms such as 'beautiful', 'art', or 'work of art',
and implies that such terms are family-resemblance concepts (LC 10; AWL
35-6; CV 24). There are no conditions which are individually necessary and
joindy sufficient for the application of these terms. Their instances are rela­
ted in a multiplicity of ways, through a 'complicated network of overlapping
similarities'. One of the arguments to this effect is fallacious. Wittgenstein
notes that terms like 'beautiful' and 'good' are bound up with what they
modify - the features which constitute beauty in a face do not do so in a
sofa. However, this does not show that 'beautiful' is a family-resemblance
concept, but only that, like 'good', it is used attributively rather than pre-
dicatively. Wittgenstein's treatment is most convincing with respect to the
terms 'art' and 'work of art'. There may be necessary conditions here: art is
a human activity and a work of art an artefact. But there is no single condi­
tion by virtue of which the artefacts of Beethoven, Beuys, Brecht, Cage,
Giotto, Jandl, Praxiteles, Pollock and Warhol qualify as works of art.
(d) Wittgenstein rejects the idea that aesthetics is a branch of psychology,
aiming to provide causal explanations of our aesthetic experience. There are
three aspects to this position, (i) Wittgenstein rejects causal accounts of artis­
tic value, in particular hedonistic theories which conceive of aesthetic value
as a tendency to cause experiences of pleasure or displeasure. He accuses
them of what later became known as the 'affective fallacy', namely of con­
fusing the value of a work of art with the psychological effect it has on
people. On such accounts, the value of a work lies in its causal effect (the
experience it produces). However, that experience might be created by other
means, through another work or even a drug. The only way of appreciating
a work of art is by experiencing and understanding its intrinsic features; its
value is determined not by any causal effects it might have, but by these
features, as measured against certain standards. The question 'Why is this
beautiful/valuable?' cannot be answered by a causal explanation (M 104-7).
(ii) Wittgenstein insists that the relationship between an aesthetic judge­
ment or impression and its object (the work) is intentional and hence inter­
nal, not an external or causal one — a specific case of his general rejection
of causal theories of INTENTIONALITY. My judgement that Vaughan Williams'
music is primitive is not about the causes of my reactions, which may be
anything from neurophysiological events to being prejudiced against English
composers. It is about those features which are mentioned in my judgement
or in my subsequent explanations of it.
(iii) Wittgenstein claims that aesthetic explanations are neither causal nor
subject to experimental checks: 'an aesthetic explanation is not a causal
explanation.' This dictum covers not just explaining why something is beau-
33
 AESTHETICS
tiful or impressive, but any explanation of 'aesthetic impressions'. 'The puz­
zles arising from the effects the arts have are not puzzles about how these
things are caused.' They are not solved by experiment, since the correct
answer is the one which satisfies the subject in question (LC 11-18, 21, 2 8 -
9). This claim seems to be refuted by stories like that of Soderini, who
claimed to be dissatisfied with the nose of the David, but had his qualms dis­
sipated after Michelangelo pretended to have altered it. However, the fact
that Soderini's dissatisfaction could be removed without any alteration to the
nose no more shows that it was about something other than the nose, than
the fact that my desire for an apple can be removed without my getting an
apple (e.g., by a punch in the stomach) shows that it was something other
than an apple that I desired (PR 64; see INTENTIONALITY). But it does show
that there is a type of aesthetic explanation which is causal, and hence
empirical, namely of what makes us react in a certain way, something of
which we may be unaware.
However, Wittgenstein's failure to take into account such aetiological
explanations does not vitiate his account of ordinary aesthetic explanations,
which specify either the object of our aesthetic reactions, or their motives or
reasons. They explain our reactions through enhancing our understanding
of the work itself. Ordinary aesthetic explanation is descriptive, in a general
sense. Apart from straightforward reference to aesthetic standards, this may
involve the following, (a) Pointing out analogies between the work under
consideration and others: we place the work under consideration side by
side with other items. Sometimes these comparisons are synaesthetic (e.g.,
when Brahms' music is elucidated by reference to Keller's novels). In other
cases it is a matter of ASPECT-PERCEPTION, bringing to notice a hitherto unno­
ticed feature of the work by placing it in a new environment or altering it in
a certain way - thus we may come to notice the power of Klopstock's
poems when read in a certain metre (LC 4, 32n; PI LI 207; RPP I §§32—7).
(b) Certain gestures may help us to understand, in particular, great works of
art, which cannot be adequately characterized by reference to aesthetic stan­
dards. In such situations we often use words in what Wittgenstein calls an
'intransitive' way. We say that the musical phrase has a particular sig­
nificance, but not as a preliminary to specifying what significance it has. This
does not mean, however, that it is ineffable. Often we manage to bring out
features of music through gestures or facial expressions, as happens, for
example, when a conductor explains a musical phrase to an orchestra
through gesticulations (BB 158, 178-9; PI §523; CV 69-70).
Wittgenstein's most important contribution to contemporary aesthetics has
been his application to aesthetics of the idea of family resemblance. Anglo­
phone aesthetics in the twentieth century has been preoccupied with the
question 'What is Art?', partly because modern art itself has self-consciously
ANTHROPOLOGY
posed this question as a challenge, and partly because the linguistic turn
initiated by Wittgenstein put such analytic questions at the heart of the phi­
losophical enterprise. The attempts to answer the question through an ana­
lytic definition have generally been viewed as unsuccessful (often they are
blatandy circular) and futile. As a result, Wittgenstein's idea of family resem­
blance has been welcomed as a liberation, an acceptance which has by and
large led to the abandonment of attempts to discover the essence of art.
anthropology Wittgenstein made two stimulating, if brief and unpol­
ished, contributions to the methodology of anthropology. One is his discus­
sion of FORMS OF LIFE and of radical translation. The other is his harsh
remarks on Frazer's Golden Bough. Frazer attempted to explain a rite in clas­
sical antiquity - the succession of the King of Nemi - by reference to simi­
lar rituals around the world. Wittgenstein raised the following objections to
Frazer's procedure (GB 118-33; AWL 33-4; M 106-7): (a) Frazer's collec­
tion of data about other rituals does not provide the genetic explanation of
the Nemi rite he sought, but rather the raw materials for an OVERVIEW
which explains why we find the ritual horrifying by linking it to basic
human impulses with which we are familiar; (b) the very attempt to provide
a genetic explanation of the rite should be abandoned in favour of describ­
ing it; (c) Frazer presents these rites as instrumental, as aiming at the bring­
ing about of certain causal consequences, and hence as based on false
empirical beliefs or on proto-science, when in fact they are expressive or
symbolic.
Objection (a) is plausible. Nothing but thin analogies and groundless con­
jectures supports Frazer's genetic account, while the similarities to and dif­
ferences from other rites do provide a non-genetic type of insight into the
nature of the Nemi rite. Objection (b) is more problematic. When Wittgen­
stein condemns genetic accounts he must have in mind not that they are
illegitimate as such, but that they must be distinguished from understanding
what ritual acts mean. Moreover, Wittgenstein does not maintain that the
only way of understanding what a ritual means is to link it to universal
human impulses or emotions. He states explicidy that explanations can
make reference to the beliefs of the participants of the ritual (GB 128). But
the historical origins of a ritual matter to its meaning only in so far as the
participants themselves attach significance to them. The eating of unleavened
bread at the Passover Feast is to be understood as an act of commemora­
tion. But what matters to this understanding is not the fact that the children
of Israel ate unleavened bread in the desert, but that pious Jews today
believe that they did. However, even if one distinguishes understanding what
a ritual means from understanding how it came about, it is implausible to
hold, as Wittgenstein does, that the latter contributes nothing to the former.
Someone who knows how the beliefs and the practices have evolved may
35
 ASPECT-PERCEPTION
well be in a better position to understand their content. Moreover, some
ceremonial acts have no expressive, symbolic or instrumental function, but
are simply performed because they accord with tradition (e.g., the pacing
back and forth of the Proctors at an Oxford degree ceremony). The only
kind of explanation of such rituals that can be given is by reference to their
origins and to a ritualistic tradition.
As regards (c), Wittgenstein is right to draw attention to the •expressive
and symbolic nature of many ritual acts. We do not burn effigies or kiss the
pictures of loved ones to achieve a certain effect. 'Magic brings a wish to
representation; it expresses a wish.' Moreover, 'if the adoption of a child
proceeds in such a way that the mother draws it from under her clothes [as
is the case among the Bosnian Turks], it is surely insane to believe that an
error is present and that she believes she has given birth to the child.' Witt­
genstein occasionally acknowledges that some rituals are instrumental. But
he also insists that all magical rituals are symbolic. However, many rituals
which we would count as magic aim at producing a certain effect, and are
based on superstition, on false beliefs in supernatural mechanisms. Wittgen­
stein seems to assume that if ritual practices were instrumental, they would
be 'sheer stupidity' (GB 119, 125). However, while superstition is irrational,
it is not simply stupid, but an expression of pervasive and profound human
fears and aspirations.
aspect-perception This term denotes a gamut of interrelated perceptual
phenomena. The paradigmatic case is what Wittgenstein calls 'aspect-dawn­
ing' or 'change of aspect' (Aufkuckten eines Aspekts or Aspektwechsel): certain
objects, especially schematic drawings - 'picture-objects' (PI II 194; LW I
§489) - can be seen under more than one aspect. An aspect dawns on us
when we notice such a hitherto unnoticed aspect of the object we are look­
ing at, come to see it as something different. Thus we may pass from seeing
a 'puzzle picture' as a mere collection of lines to seeing it as a face; from
seeing Jastrow's 'duck-rabbit' as the picture of a duck to seeing it as the pic­
ture of a rabbit.
Puzzle pictures like the Necker cube briefly appear in the early work.
From 1935 onwards, Wittgenstein's philosophy of psychology returns time
and again to seeing-as (TLP 5.5423; NB 9.11.14; BB 162-79; PI II 193-229;
36
ASPECT-PERCEPTION
RPP I & II passim; LW I passim; LW II 12-17). Between 1947 and 1949, it
dominates his work, partly under the influence of Kohler's Gestalt psychol­
ogy. Wittgenstein's immediate aim was to dissolve the paradoxical appear­
ance of aspect-dawning: when looking at a picture-object we can come to
see it differendy, although we also see that the object itself remains unchan­
ged. It seems to have changed and yet seems not to have changed (PI II
193-5; LW I §493).
One way of dealing with aspect-dawning is to point out that perceiving
alternative aspects is caused by different patterns of eyeball movements.
Wittgenstein was aware of such correlations, but denied that they resolve die
paradox (PI II 193, 203, 212-16; LW I §795). For even if they explain why
the phenomenon occurs, they do not provide a description of it which
escapes the paradox. Wittgenstein attached enormous importance to aspect-
perception, since he thought that in this phenomenon 'problems about the
concept of seeing come to a head' (LW I §172). Presumably this is because
it exemplifies in a precise form the concept-saturatedness of perception. We
see one and the same thing (e.g., a person's face) but may see it differendy
(e.g., as placid or anguished).
Wittgenstein's discussion concentrates mainly on Gestalt psychology.
According to Kohler, what we perceive immediately is not a mosaic of dis­
crete and unorganized stimuli (dots and coloured surfaces, sounds), as
empiricism and behaviourism have it, but Gestalten, circumscribed and orga­
nized units, such as material objects or groups of objects (Gestalt ch.V). We
do not see three dots, but see them form a triangle; we do not hear a chao­
tic array of sounds, but detect a melody. This is close to Wittgenstein in
rejecting the reductionist view according to which we construct perceptual
objects out of raw data. Unfortunately, Kohler's treatment of aspect-dawn­
ing reifies Gestalten. He claims that in aspect-perception we do riot see one
and the same object under different aspects, but rather two different 'visual
objects' or 'visual realities' (Gestalt 82, 107, 148-53). In his attempt to do jus­
tice to the idea that we see the picture-object differendy, Kohler thus turns
an aspect (Gestalt) into a private mental entity. This reification is not just ter­
minological, it is integral to his account of aspect-perception. The two
'visual objects' are said to differ in their organization, which is as much a
feature of them as their colour and shape. Accordingly, what changes in
aspect-perception is not the colour or shape of the elements of the visual
impression, but their organization. Wittgenstein rejects this explanation. It
suggests that what changes in cases of aspect-dawning is the way we per­
ceive the (spatial) relationships between the elements of the picture. But this
is mistaken. When asked to depict faithfully what we see, that is, the pic­
ture-object, before and after an aspect-change, there is no more difference
in the organization of the elements than in their shape or colour (although
we may go about depicting the object differendy). The characteristic of
37
ASPECT-PERCEPTION
aspect-dawning is precisely that no specific feature of the visual field chan­
ges. The alleged change in organization cannot be specified, which means
that Kohler's 'organization' could only refer to ineffable features of a private
object, which the PRIVATE LANGUAGE ARGUMENT excludes as chimerical (PI IT
196-7; RPP I §§536, 1113-25; LW I §§444-5, 510-12).
An alternative to the 'Gestalt' explanation is that what has changed is our
interpretation, not of a private impression, but of the object perceived. This
raises a question which dominates Wittgenstein's discussion, namely whether
noticing an aspect is a case of seeing or of thinking. Wittgenstein's verdict
on this issue is ambivalent. His first point is that types of aspect-perception
differ according to the degree of thinking involved (PI II 207-12; LW I
§§179, 530, 582-8, 699-704; RPP I §§1, 70-4, 970; RPP II §§496, 509). At
one end lie 'conceptual' aspects like those of the duck-rabbit, which cannot
be expressed solely by pointing to parts of the picture-object, but require
possession of the relevant concepts. At the other, lie 'purely optical' cases,
such as the 'double cross' in which we can express our seeing of the
aspect by following certain lines of the picture-object, without using concepts
(but even here concepts like background and foreground seem involved).
His second point is that the concept of seeing an aspect lies between that
of seeing, which is a state, and that of interpreting, which is an action. It is
closer to the latter in the following respects (PI II 212; RPP I §§27, 169;
RPP II §§544-5; LW I §§451, 488, 612):
the 'optical' or 'visual' picture remains the same, as we have seen;
aspect-seeing, unlike most cases of seeing, is subject to the will: although
we may not always succeed in noticing an aspect or keeping it in focus,
it always makes sense to try to do so, and we often succeed;
in noticing an aspect of the conceptual kind we do not just focus on
properties of the object perceived, but realize certain INTERNAL RELATIONS
between it and other objects, relations of similarity and dissimilarity
such as those between two human faces.
Aspect-perception is closer to seeing in the following respects (PI II 203-4,
212; RPP I §§8, 1025; RPP II §§388, 547):
there is no possibility of being mistaken about seeing an aspect;
aspect-seeing is a state; in particular it has 'genuine duration', that is, it
has a beginning and an end which can be clocked, can be interrupted,
etc.;
there is no more direct expression of the experience than the report of
aspect-perception 'I see it as a rabbit', that is, there is no sharp contrast
between the 'interpretation' and the uninterpreted data.
38
ASPECT-PERCEPTION
It may seem that Wittgenstein creates an artificially stark contrast between
seeing and thinking by restricting the latter to interpreting, to conjecturing
what a picture represents (RPP I §§8-9, 13, 20; RPP II §390; PI II 193,
197, 212; LW II 14). But the paradox of aspect-dawning does not depend
on such a narrow conception of thinking. I may know about the duck-rabbit
picture, and think 'One can see a rabbit here', without being able to see it.
Wittgenstein suggests that the paradox trades on an equivocation: what I
see in the ordinary sense has not changed, while what I see in the sense of
'seeing' closer to thinking has. Given the extent to which he has laboured
the paradox, this solution is more a whimper than a bang. Nevertheless, it
features important insights. Reports of aspect-dawning are not descriptions,
either direct or indirect (interpreting), of an inner experience which accom­
panies ordinary perception, but AVOWALS, spontaneous reactions to what we
see. Moreover, what changes in aspect-dawning is not what we perceive, or
its 'organization', but our attitude to it, how we react to it and what we can
do with it. Suddenly, we copy or explain the puzzle picture differendy,
change the way we play a piece of music or recite a poem (PI II 197-8,
208; RPP I §982; LC 1-11). One important thing we do in noticing an
aspect is placing what we perceive in another context we detect new con­
nections or draw fresh comparisons. This is why changing the context of an
object may alter our perception of it (PI -II 212; RPP I §1030; LW I §516).
Wittgenstein illustrates aspect-seeing through 'aspect-blindness', the inabil­
ity to experience aspect-dawning (PI II 213-14; RPP II §§42, 478-9, 490;
LW I §§492-3, 778-84). An aspect-blind person could apply a new descrip­
tion to a picture-object, use, for example, the schematic drawing of a cube
as a picture of a three-dimensional object. But he would not experience this
as seeing something differendy, experiencing a jump in aspect, and would
not recognize the incompatibility with treating it as a two-dimensional com­
plex of three parallelograms. His defect is one not of sight but of imagina­
tion.
A special kind of aspect-blindness is 'meaning-blindness', the inability to
experience the meaning of a word (PI II 175-6, 210; RPP I §§189, 202-6,
243-50, 342-4; cp. James's 'soul blindness', Psychology I ch. II). This does
not reinstate the idea that the meaning of a word is a mental phenomenon
which accompanies UNDERSTANDING. Instead, Wittgenstein claims that words
have a 'familiar physiognomy': they are associated with other words, situa­
tions and experiences, and can assimilate these connections. Thus one may
feel that names 'fit' their bearers. And words turn into mere sounds if these
connections are severed, e.g. when they are mechanically repeated several
times (PI II 214-15, 218).
Experiencing meaning underlies the 'secondary sense' of terms: some
people are inclined to say things like ' " e " is yellow', ' " u " is darker than
" i " ' or even 'Tuesdays are lean, Wednesdays are fat.' This secondary sense
39
ASPECT-PERCEPTION
(a) differs from the primary sense: obviously 'e' is not yellow in the sense in
which flowers are - it cannot be compared with a sample of yellow; (b) pre­
supposes the primary sense: it can only be explained by reference to the pri­
mary one, but not vice versa; (c) is not a matter of ambiguity or of
metaphor: we can disambiguate 'bank' by introducing a new term and can
paraphrase metaphors, but we cannot express secondary senses in any other
way (PI II 216; LW I §§797—8). Secondary sense also explains sylleptic
ambiguity: the fact that we speak of deep sorrows and wells, of plaintive
cries and melodies, etc.
The meaning-blind person uses and explains words correcdy, but has no
'feel' for their physiognomy, a lack which is comparable to the lack of a
musical ear. For this reason, he is barred from important forms of AESTHETIC
discourse, or from understanding puns.
Some passages of Wittgenstein declare that the importance of aspect-per­
ception lies in its connection with experiences of meaning, while others insist,
correcdy, that it is not essential to the concept of meaning (PI II 214; LW I
§784 vs. RPP I §358; RPP II §§242-6). A tempting explanation of Wittgen­
stein's obsession with the topic is that for him aspect-perception is integral to
all perception. This seems to be supported by his distinction between aspect-
dawning and 'the continuous seeing of an aspect'. Yet Wittgenstein denied
that seeing-as is typical of all experience. Seeing-as requires a contrast
between two different ways of perceiving an object, but under normal cir­
cumstances it makes no sense to say that, for example, one sees the cudery as
a knife and fork (PI II 194—5). Accordingly, Wittgenstein confines continuous
aspect-perception to objects like pictures. Here no special circumstances are
needed for a contrast between relating to what is perceived either as the
depiction of something else, or as an object in its own right.
It is precisely by denying that all perception is aspect-perception that
Wittgenstein rejects the empiricist myth of the given, the idea that what we
perceive immediately are raw stimuli, which we then interpret as something
eke (RPP I §§1101-2; Z §§223-5). Typically, we do not just hear noises, but
words and melodies, do not just see colours and shapes, but material
objects, not just bare bodily movement, but human BEHAVIOUR infused with
attitudes and emotions. What is constitutive of ordinary perception is that
aspect-perception is possible, under special circumstances we react to words
as sounds, or human behaviour as mere bodily movement. But while it is
always possible to describe what one perceives in terms of sounds, or colour
and shape, it does not follow that any other description is indirect or
inferred. On the contrary, it is easier to describe a person's face as 'sad',
'radiant' or 'bored', than to describe it in physical terms. We know the con­
clusions of the alleged inference, not its premises. Neural stimuli may feature
in a causal explanation of perception and understanding, but are not raw
data from which we construct objects or linguistic meaning.
40
AUGUSTINIAN PICTURE OF LANGUAGE
assertion see BELIEF
Augustinian picture of language Philosophical Investigations starts with a
quotation from the Confessions (1/8) in which Augustine describes how he
learned language as a child. Wittgenstein first mentions this passage in the
'Big Typescript' (BT 25-7; see PG 57). From the Brown Book onwards he
used it as the starting-point of what was to become the Investigations. This
marks a break in the manner of presentation of his later work. It demands
explanation, since the passage is part of Augustine's autobiography, not of
his reflections on language. The reason Wittgenstein gave for using the quo­
tation is that it stems from a great and clear thinker, and thus displays the
importance of what he refers to as 'Augustine's conception of or 'descrip­
tion of language' (PI §§1-4; EPB 117). This suggests that he treated Augusti­
ne's view not as a full-blown theory of language, but as a proto-theoretic
paradigm or 'picture' which deserves critical attention because it tacitiy
underlies sophisticated philosophical theories. The claims which Investigations
§1 extracts from the passage are:
(a) every individual word has 'a meaning';
(b) all words are names, i.e. stand for objects;
(c) the meaning of a word is the object it stands for;
(d) the connection between words (names) and their meanings (referents)
is established by ostensive definition, which establishes a mental asso­
ciation between word and object;
(e) sentences are combinations of names.
Two consequences are spelled out subsequendy:
(f) the sole function of language is to represent reality: words refer, sen­
tences describe (PI §§21-7);
(g) the child can establish the association between word and object only
through thinking, which means that it must already possess a private
language, in order to learn the public one (PI §32; see PRIVATE LAN­
GUAGE ARGUMENT).
Accordingly, the Augustinian picture comprises four positions: a referential
conception of word-meaning; a descriptivist conception of sentences; the
idea that OSTENSIVE DEFINITION provides the foundations of language; and the
idea that a language of THOUGHT underlies our public languages.
Wittgenstein was the first to subject this position to sustained criticism.
One of his strategies in Investigations §§1-64 is the use of fictional LANGUAGE-
GAMES, invented forms of communication. Thus, the language of the build­
ers (PI §§2, 6, 8) seeks to display the Augustinian picture as 'a primitive
idea of how language functions' or an 'idea of a language more primitive
41
AUGUSTINIAN PICTURE OF LANGUAGE
than ours'. Even that, however, is too generous, as can be seen from his
other objections (matched here to the claims listed above): (a) There are
'syncategorematic expressions' (articles, demonstrative pronouns, connectives
like 'if ... then') which are meaningful only within a context, (b) The
Augustinian claim is modelled solely on proper names, mass nouns and
sortal nouns. It ignores verbs, adjectives, adverbs, connectives, prepositions,
indexicals and exclamations (PG 56; BB 77; PI §27). (c) Even in the case of
noun-phrases which can be said to name or stand for something, one must
distinguish between their meaning and what they stand for. 'When Mr
N.N. dies one says that the bearer of the name dies, not that the meaning
dies' (PI §40). There are two parts to this objection: (i) if referential failure
due to the referent's ceasing to exist rendered a referring expression mean­
ingless, propositions like 'Mr N.N. died' could not make sense; (ii) identify­
ing the meaning of a word with its referent is a category mistake, namely
of confusing what a word stands for with its meaning: the referent of 'Mr
N.N.' can die, but not its meaning (AWL 44). (e) One must distinguish a
proposition like 'Plato was the pupil of Socrates and the teacher of Aris-
totie' from a mere list of names like 'Socrates, Plato, Anstotie'; only the
former says something, and thereby makes a 'move in the language-game'
(PI §22). (f) The Augustinian picture runs counter to the 'multiplicity of
language-games'. In addition to describing there are not just questions and
commands but 'coundess' other kinds (e.g., telling jokes, thanking, cursing,
greeting, praying) (PI §23). Nor is describing the highest common factor of
these various linguistic activities.
Some have held that the Augustinian picture is an all-pervasive philoso­
phical illusion, the principal target not just of Wittgenstein's philosophy of
language but also of his philosophy of psychology and of mathematics.
Others have maintained that it is too implausible to be his major target, or
the source of so many philosophical positions. Wittgenstein nowhere sug­
gests that the Augustinian picture is the only source of philosophical confu­
sion; but he maintains that complex philosophical edifices are often based
on simple pictures or assumptions. In fact, the referential conception of
word-meaning has played a prominent role in semantics since Plato. It is
not confined to the absurd suggestion that all words are proper names
which have material objects for their meanings, but includes the very idea
of 'meanings', entities correlated with the sign (as in the scholastic tag
'unum nomen, unum nominatum'). Wittgenstein accuses even nominalism
of being committed to the Augustinian picture, because it accepts that all
words either name something or name nothing, and settles for the second
alternative, in order to avoid commitment to abstract entities. The Augusti­
nian picture may also grant that there are distinct types of expressions,
while insisting that they all stand for or signify something, and that the dif­
ferences are simply due to the differences between the types of objects sig-
42
AUGUSTINIAN PICTURE OF LANGUAGE
nified. By the same token it may insist that the differences between various
uses of sentences are due to their describing different types of facts (PI
§§24, 383; PLP 143, 407).
These elaborations of the referential conception lie behind mentalist and
Platonist conceptions of meaning, which postulate non-material entities to
play the role of meanings. The mentalist version goes back to Aristode, and
influences modern linguistics via de Saussure's distinction between signifiant
and signifie. It has been rampant in British empiricism ever since Locke
claimed that all words have their meaning in virtue of standing for ideas.
Russell's theories of meaning are variations on this mentalist theme. 'Words
all have meaning in the simple sense that they are symbols which stand for
something other than themselves' (Principles 47). Russell moved away from
this extreme version of the Augustinian picture in the theory of descriptions:
expressions like 'the present King of France' are analysed as 'incomplete
symbols' which do not refer to an object. At the same time, he condones
claim (e): fully analysed propositions are combinations of 'logically proper
names' - names which stand for objects that could not fail to exist, and
hence are immune to referential failure. According to his 'principle of
acquaintance', these names are demonstratives like 'this' which refer to
sense-data. Throughout his career, Russell maintained that words have
meaning by virtue of an ostensive association with private contents of
experience. Moreover, even when, under Wittgenstein's impact, he ceased to
regard sentences as combinations of names, he remained committed to the
idea that the facts which sentences express are 'complexes', that is, con­
catenations of simple objects (Logic 200-3; Problems 79-80; Principia i.43).
Russell never abandoned the Augustinian picture. Instead he provided it
with a line of defence: although the surface of language may not correspond
to the picture, its ultimate elements, to be revealed by LOGICAL ANALYSIS, do.
The Platonist idea that meanings are not private ideas but abstract enti­
ties beyond space and time is prominent in Bolzano, Meinong and Frege.
Frege diverges from the Augustinian picture in three respects. Firstiy, he
sharply distinguishes between 'proper names' ('the morning star'), and con­
cept-words ('is a planet'). Secondly, according to Frege's 'context-principle' a
word has a meaning only in the context of a sentence (Foundations §§60-2,
106). This overcomes the semantic atomism of claim (a): a sentence can be
meaningful without every individual word's being associated with a material
or mental entity. That numerals have a meaning (which is an abstract
object) is evident from the contribution they make to the truth-values of sen­
tences in which they occur. Thirdly, Frege distinguishes between the sense
(Sinn) and the meaning (Bedeutung) of expressions, i.e. their referent ('Sense';
Laws I §2).
This two-tier model of meaning, familiar from Mill's distinction between
connotation and denotation, avoids the problem of referential failure without
43
 AUGUSTINIAN PICTURE OF LANGUAGE
postulating logically proper names, since an expression without a 'meaning'
can have a sense. But in other respects it remains wedded to the Augusti­
nian picture. Frege's dichotomy incorporates claim (c): 'the word "meaning"
is being used ilhcidy', namely for the thing that corresponds to the word (PI
§40). Moreover, in his ideal language every expression (apart from the asser­
tion-sign) not only expresses a sense but refers to a meaning. Worse, 'senses'
are themselves abstract entities inhabiting a Platonic 'third realm' ('Thought'
68-9). In this respect they merely add to the number of entities the Augusti­
nian picture assigns to words. Finally, although concept-words differ from
proper names, they are still names, namely of abstract entities (functions); so
are mathematical and logical symbols, and even sentences, which name
either one of two 'logical objects', the True and the False.
The Tractatus moves further away from the Augustinian paradigm. It
rejects both the idea that LOGICAL CONSTANTS (propositional connectives,
quantifiers) are names of entities, and the ensuing view that the propositions
of LOGIC are descriptions of some kind of reality. It also insists that a PRO­
POSITION is not a name of anything, but a sentence-in-use, a propositional
sign in its projective relation to the world. Saying is not naming. Using Fre­
ge's distinction, Wittgenstein claims that only propositions have a 'sense',
and only names have a 'meaning'. At the same time, the Tractatus holds that
all constituents of fully analysed propositions are names. The elementary
propositions of which complex propositions are formed are 'a nexus, a con­
catenation, of names' (TLP 3.201f., 4.22f). It tries to avoid claim (e) by
insisting that propositions, unlike lists of names, are FACTS: they have a struc­
ture (LOGICAL FORM), which, along with the meaning of their constituent
names, determines their sense. However, like Frege, the Tractatus explicidy
condones claim (c): 'A name means an object. The object is its meaning'
(TLP 3.203). Finally, the PICTURE THEORY is based on the idea that the only
meaningful propositions are those that describe possible states of affairs.
All this suggests that many august semantic theories lie in the target-area
of the Investigations' attack. That attack is completed by Wittgenstein's alter­
native: the meaning of a word is its USE; for some expressions, that use is to
refer to an object, and they can be explained by pointing at their referent
(PI §43). Many critics of this alternative revert to elements of the Augusti­
nian picture. Thus it has been claimed that what matters about the use of a
word so far as its meaning is concerned, is precisely what it stands for or
signifies. Furthermore, the axioms of contemporary truth-conditional seman­
tics correlate singular terms with objects, and predicates with ordered sets of
objects. Finally, it is generally accepted that although the Investigations righdy
insisted against the Tractatus that there are different types of speech acts,
even an imperative or a question contains a descriptive element, its sense (a
thought or assumption), which must be distinguished from its 'force'.
In his discussion of BELIEF, the later Wittgenstein challenges this sense/
44
AUTONOMY OF LANGUAGE
force distinction. Whether or not his attacks are successful, the Augustinian
picture is not a straw-man. Equally, however, there is no 'guilt by associa­
tion': a semantic theory cannot be attacked simply for containing 'Augusti­
nian ideas', since there are important connections between meaning and
reference. This also goes for the influence of the Augustinian picture in
areas other than language. Reification is a paradigmatic 'Augustinian' temp­
tation. If all nouns are names, psychological expressions must name mental
objects, events, processes or states; mathematical and logical terms must
name abstract entities. Platonism and Cartesianism postulate separate onto-
logical realms which are inhabited by the alleged referents of abstract nouns
like numerals or of mental terms like 'pain' and 'understanding'. There is no
doubt that this move is a recurrent target of Wittgenstein's philosophy of
mathematics and of psychology. He also challenges the seemingly self-evi­
dent claim that mathematical propositions and first-person psychological
utterances 'describe' abstract or mental objects. At the same time, the attack
on the Augustinian picture does not finish off these venerable positions. It
shows only that it is misguided to insist that words must refer, or sentences
describe, not that the expressions at issue actually fail to do so. Moreover,
while the Augustinian picture is one possible reason for adopting these posi­
tions, there are other, and often stronger reasons, such as the objectivity and
inexorability of mathematics, or the indubitability of AVOWALS.
autonomy of language, or arbitrariness of grammar These terms
indicate the idea that GRAMMAR, the linguistic rules which constitute our con­
ceptual scheme, is arbitrary, in the sense that it does not pay heed to any
putative essence or form of reality and cannot be correct or incorrect in a
philosophically relevant way. This provocative claim is directed against lin­
guistic foundationalism, the view that language should mirror the essence of
the world. One version of this is the search for an ideal language, like that
of Leibniz, Frege and Russell, which is supposed to mirror the structure of
thought and reality more accurately than ordinary language (Posthumous 266;
Logic 185-234, 338).
The Tractatus rejects the idea that natural language could be logically
flawed, but embraces an alternative version of linguistic foundationalism.
Any language capable of depicting reality must be governed by LOGICAL
SYNTAX, which is a 'mirror-image of the world' (TLP 6.13). Its rules must
match the structural features of reality: the LOGICAL FORM of names must
mirror the essence of the objects they stand for. At the same time, 'logic
must take care of itself (TLP 5.473). The SAYING/SHOWING distinction prohi­
bits a doctrine like Russell's theory of types which justifies logical syntax by
reference to reality: any proposition that purports to justify logical syntax
must be meaningful, and hence presupposes logical syntax. Yet, the extra-
linguistic foundations of logic show themselves, in the logical form of elemen-
45
AUTONOMY OF LANGUAGE
tary propositions and in the fact that certain combinations of signs are
TAUTOLOGIES (TLP 6.124). Finally, while the superficial features which distin­
guish different languages are arbitrary, there is only one 'all-embracing logic
which mirrors the world', common to all sign-systems capable of picturing
reality (TLP 5.511; see NM 108-9).
That the Tractatus contains a foundationalist 'mythology of symbolism'
(PG 56; Z §211) is confirmed by subsequent comments. After 1929, Witt­
genstein initially insisted that unlike games, grammar is not 'arbitrary',
because it has to mirror the multiplicity of facts; and that 'the essence of
language is a picture of the essence of the world', albeit not in propositions,
but in grammatical rules (LWL 8-10; PR 85). Gradually he came to hold
that the apparent essence of reality is nothing but a 'shadow of grammar'.
Grammar constitutes our FORM OF REPRESENTATION, it determines what counts
as a representation of reality, and is not itself responsible to reality (PG 88,
184; PI §§371-3). There are three major aspects to this autonomy.
(a) Grammar is self-contained, not responsible to extralinguistic reality, (i)
Wittgenstein attacks the idea that there is a MEANING-BODY behind a sign, a
non-linguistic entity — its meaning — which determines how it can be used
correcdy. Grammatical rules do not somehow follow from 'meanings', they
partly constitute them. Signs as such don't have meanings; we give them
meaning by adopting certain standards of linguistic correctness, by explain­
ing and using them in a certain way (BB 27-8).
(ii) There is a natural view, explicit in empiricism and implicit in the
Tractatus, that OSTENSIVE DEFINITIONS forge a link between a word and its
extralinguistic meaning, thus grounding language in reality. Wittgenstein
now argues that the samples used in ostensive definitions are part of gram­
mar, in that they function as standards for the correct application of words,
as do colour samples for colour-words.
(iii) Wittgenstein rejects the idea that the rules of LOGICAL INFERENCE can be
justified either by empirical facts or through model-theoretic proofs.
(iv) A powerful challenge to the idea of the self-containedness of grammar
is the Lockean idea of 'real essences', revived by Kripke and Putnam. When
we found that certain substances that used to be called 'gold' because they
satisfied superficial criteria have a different atomic structure from gold, we
did not conclude that gold does not always have the atomic number 79, but
distinguished between real gold and, for example, fool's gold. Consequendy,
the real meaning of words is determined not by the rules we adopt, but by
the 'real nature' of the things referred to, which science discovers. Wittgen­
stein anticipated this line of argument. We sometimes change the CRITERIA
for the applications of words. But this amounts to conceptual change
sparked off by an empirical discovery, not to a discovery of 'the real mean­
ing' (Z §438). Putnam objects that this ignores the fact that we now know
46
AUTONOMY OF LANGUAGE
more about gold than before. Wittgenstein could reply that we know more
about gold, that is, about the atomic consistency of a certain stuff, without
knowing more about the meaning of 'gold'. The latter is determined by our
EXPLANATION of meaning, which specifies criteria that must be fulfilled by
anything we call 'gold'. And we distinguish between UNDERSTANDING the term
and having expert chemical knowledge. But even if science does not dis­
cover meanings, we, for good reasons, change certain concepts in accor­
dance with its findings, and to this extent language is not autonomous. One
might further claim that the new concept is simply correct, since it corre­
sponds to objective features of a stuff (gold). However, that stuff has an
indefinite number of objective properties. These could all be used to define
different concepts, which may be more or less useful, or have more or less
explanatory power. But that is not a matter of corresponding to reality.
(b) Grammatical rules cannot be justified. Even if grammatical rules cannot
be justified by reference to reality, might they not be justified in the same
way as strategic or technical rules, by reference to their purpose or function?
Wittgenstein resists this (PG 184-5, 190-4; PI §§491-6; Z §§320-2; MS 165
106; BT 194—5). We can justify the rules of an activity like cooking by refer­
ence to its goal, with cooking the production of tasty food, since that goal
can be specified independendy of the means by which it is attained. But we
cannot justify the rules of language by reference to a goal like communica­
tion, since the relationship between language and communication is con­
ceptual, not instrumental. A sound-system which does not fulfil the purpose
of communicating is not a worse language, but no language at all. (Note,
however, that this sits uneasily with Wittgenstein's simultaneous insistence
that language cannot be defined as a means of communication since it is a
FAMILY-RESEMBLANCE concept.)
Wittgenstein also provides a quasi-Kantian argument against any attempt
to justify grammar by reference to facts. We cannot invoke facts in support
of grammatical rules without expressing them in language. Hence, to justify
a grammatical rule could only mean to support it by adducing a PROPO­
SITION. But any such proposition is expressed in some language, and there­
fore presupposes a certain grammatical framework. There is no such thing
as an extralinguistic or preconceptual perspective outside any grammar
from which we could justify a given grammatical system (see TRUTH). This
confronts the foundationalist with a dilemma. Either the grammar of the
supporting proposition is identical with that of the rule to be justified. In
that case the justification is circular. Or the supporting sentence belongs to
a different grammatical system. This would avoid circularity, but only at
the cost of incommensurability. A different grammatical system defines dif­
ferent concepts, hence a statement in a different system can neither justify
nor refute grammatical propositions of our system. We cannot justify the
47
 AUTONOMY OF LANGUAGE
grammar of our colour-words by claiming that there are precisely four pri­
mary colours which objectively resemble each other, because the concept of
similarity upon which this move depends is part of the grammar we seek to
justify. The foundationalist could only provide a conceptually independent
justification, and thus avoid the first horn of the dilemma, if he could
allude to the possibility of a fifth primary colour and deny that this possibi­
lity is realized. But this lands him upon the second horn, since the possibi­
lity of a fifth colour is precisely excluded as nonsensical by our rules. Each
form of representation creates its own concepts and thus lays down its own
standards of what it makes sense to say, which means that justification and
what is to be justified would pass each other by (PR 54-5; PG 97, 114;
LWL 83).
(c) Alternative forms of representation are not irrational in an absolute sense.
It seems obvious that certain essential features of language are superior to
any genuine alternatives. Wittgenstein rejects even this modest suggestion, by
reference to various alternative norms of representation (e.g., deviant ways of
counting, calculating and measuring). 'One symbolism is in fact as good as
the next; no one symbolism is necessary' (AWL 22, see 63, 117; RFM 38,
91-4, 105-6; LFM 201-2; RR 121-2). The rationale for this view is that
every form of representation provides a framework for dealing with 'recalci­
trant' experiences without having to surrender the form of representation
itself (AWL 16, 39-40, 70). Prefiguring Kuhn's idea of a scientific paradigm,
Wittgenstein illustrates how one could hold on to Newton's first law of
motion come what may. If a body does not rest or move with a constant
motion along a straight line, we postulate that some mass, visible or invisible,
acts upon it.
Alternative forms of representation are possible even in mathematics. It is
possible to adopt '12 x 12= 143' as a norm of representation (LFM 97). It
has been objected that a community which did so would have to count in a
manner which its members would recognize as mistaken. But to say that
they must have made a mistake is to adopt our norm of representation
'12 x 12= 144'. For them, by contrast, something must have gone wrong
when they count 144 objects. This may appear unconvincing: when these
people count twelve groups of twelve objects they will get 143 only by leav­
ing an object out. However, they could hold on to their norm of representa­
tion without appearing to themselves to have committed such a mistake, by
assuming that things arranged in twelve groups of twelve increase in
number by one whenever they are counted. Moreover, their allegiance to
their form of representation does not differ in kind from our allegiance to
our own. If it turned out that whenever we count twelve groups of twelve
things we get 143, we would not abandon '12 x 12=144', but look for
explanations elsewhere. However, such ad hoc assumptions would not work
48
AUTONOMY OF LANGUAGE
for numbers we can count at a glance, and this is one factor which restricts
the possibility of alternative forms of representation.
According to a naturalist interpretation, Wittgenstein's alternative techni­
ques are not meant to be intelligible, but meant to illustrate that it is a con­
tingent fact that we speak and act as we do. Wittgenstein himself, however,
claimed that divergent concepts become 'intelligible' if we imagine 'certain
very general facts of nature to be different' (PI II 230; RPP I §48; RFM 91,
95). Indeed, some of his examples are no less intelligible than the medieval
practice of measuring by the ell. What is unintelligible according to Wittgen­
stein is only the idea of changing our form of representation while retaining
our present concepts. But this reply seems to confront a dilemma. Either
alternative techniques make for different concepts, in which case Wittgen­
stein is not entided to speak of alternative forms of, for example, measuring.
Or the alternative technique counts as a form of measuring because it
shares with our techniques a certain function (e.g., of allowing the fitting
together of building-blocks), in which case our techniques are clearly super­
ior. To this Wittgenstein would reply that such functional constraints are
themselves conditional on certain needs and interests. Alternative techniques
may be inferior as means of achieving our ends. But a pre-technological
community which is only interested in measuring cloth can get by with ells,
irrespective of the fact that the lengths of people's arms vary. The basis of
calling this a form of 'measuring' lies in the fact that it plays an analogous
role in their form of life.
Nevertheless, Wittgenstein acknowledges that there are limits to revising
our form of representation. On the one hand these are conceptual. While
our concepts of counting, measuring, etc. are flexible enough to accom­
modate certain variations, there is a much tighter link between the 'laws of
logic' and notions like 'reasoning', 'thinking' and even 'proposition' or 'lan­
guage' (RFM 80, 89-95, 336; LFM 201-2, 214). A practice which does
not conform to the rule for the modus ponens simply does not qualify as
inferring. And a system which allows the derivation of a contradiction does
not count as an alternative logic. However, this does not jeopardize the
autonomy of grammar. For these limits are set not by Platonic entities, as
Frege had it, or by a 'METALOGICAL' obligation to avoid contradictions, as
the logical positivists thought, but by our concepts, by what we call 'infer­
ring', 'reasoning', or. '(a system of) rules' (PG 111, 304; WVC 199-200;
AWL 4). And the rules for the use of these terms pay no more heed to
reality than those of other words; rather, a practice which does not con­
form to them would be unintelligible to us, and would not count as a lan­
guage (note the parallels with Davidson's argument against the idea of an
untranslatable language).
There are also pragmatic constraints. Norms of representation cannot be
metaphysically correct or incorrect. But given certain facts - biological and
49
AVOWAL
socio-historical facts about us and general regularities in the world around
us - adopting certain rules can be 'practical' or 'impractical' (AWL 70). Pro­
vided that the world is as it is, people who employed alternative scientific
paradigms, ways of calculating or measuring for purposes similar to ours,
would have to make adjustments which would eventually collapse under
their own weight. Drastic changes in certain facts could render certain rules
not only impractical but even inapplicable (RFM 51-2, 200; RPP II §§347-
9; see FRAMEWORK).
The autonomy of language does not amount to an 'anything goes' relati­
vism. Grammar is not arbitrary in the sense of being irrelevant, discre­
tionary, easily alterable or a matter of individual choice. Language is
embedded in a FORM OF LIFE, and is hence subject to the same restrictions as
human activities in general. The idea of the autonomy of grammar is provo­
cative. Yet its ultimate rationale is a grammatical reminder: we call proposi­
tions true or false, but not concepts, rules or explanations. A unit of
measurement is not correct or incorrect in the way that a statement of
length is. Grammatical rules can be correct in the sense of conforming to an
established practice, or of serving certain purposes. But Wittgenstein has
made out a powerful case against the idea that they have to mirror a puta­
tive essence of reality.
avowal This term was introduced into philosophy by Ryle, but it is also a
common translation of Wittgenstein's Aufierung or Ausdruck (alternatives being
'expression', 'manifestation', 'utterance'). Wittgenstein characterized some
uses of first-person present tense psychological sentences as avowals. Nega­
tively, this indicates that they are not descriptions or reports of private
mental entities encountered in an inner realm. Positively, Wittgenstein char­
acterizes avowals as expressive in the way in which a gesture or frown
expresses or manifests emotions, attitudes, etc. They are partial substitutes
for, and learnt extensions of, natural expressions of the mental, such as cries,
smiles or grimaces. Sensation-words 'are connected with the primitive, the
natural, expressions of the sensation and used in their place. A child has
hurt himself and he cries; and then adults talk to him and teach him excla­
mations and, later, sentences. They teach the child new pain-behaviour' (PI
§244). This is not armchair learning-theory but a claim that logically the
function of avowals is akin to that of non-verbal manifestations.
This idea is crucial to Wittgenstein's rejection of the INNER/OUTER picture,
and developed out of his break with the PICTURE THEORY, according to which
all meaningful propositions express a thought and represent how things are.
The view that all meaningful propositions are descriptive survived in the
VERFICATIONISM of the transition period, which insists that a proposition that
cannot be conclusively verified lacks sense. Wittgenstein concluded that only
50
AVOWAL
sense-datum statements which describe immediate experience are genuine
PROPOSITIONS: only they allow of conclusive verification by being direcdy
confronted with experience. Accordingly, verification means something dif­
ferent in the first-person case
(1) I am in pain
from what it means in the third-person case
(1') N.N. is in pain
which is verified by reference to N.N.'s BEHAVIOUR.
In 1932, Wittgenstein realized that verification applies only to cases like
(1'), which is verified by reference to behavioural CRITERIA, not to cases like
(1). There are no intelligible answers to the question 'How do you know
that you have a pain?' (M 98-9; LSD 13; Z §436). (a) 'Because I feel it' will
not do, since there is no difference between feeling a pain and having a
pain. For one cannot have a pain and not feel it, or feel a pain one does
not have. Consequendy, the answer amounts to 'I know I have a pain
because I have it', which is vacuous, (b) 'By introspection' presupposes that
one can 'look to see whether one has it', which does not make sense, for
there is no such thing as perceiving or misperceiving one's pain.
As a result Wittgenstein detects a fundamental difference between psycho­
logical and other predicates. There is rough logical parity between
(2) I weigh over 100 kg
and
(2') H.G. weighs over 100 kg.
By contrast, there is a logical asymmetry between (1), which is an avowal,
and (1'), which is a description. Unlike descriptions, avowals: (a) do not
allow of verification, for there is no such thing as my 'finding out' that I
have a sensation or intend to go to London, or of my 'perceiving' or
'recognizing' my sensations or experiences; (b) do not allow of significant
error, ignorance or doubt; there is no room for misidentifying their subject
(see I/SELF) or misapplying their predicates: 'I thought I had a pain, but it
turned out to be an itch, and it was Sarah's, not mine' is nonsense; (c) do
not express knowledge claims (Z §§472, 549; PI §§290, 571; LPE 319; see
PRIVACY).
Occasionally Wittgenstein suggests that avowals are not cognitive because
they are not descriptions; sometimes he intimates that they are not descrip­
tions because they do not express knowledge. Ultimately, both claims are
based on the idea that there is a grammatical link between epistemic con­
cepts and the concept of description (RPP I §572; Z §549; LW I §51). Genu­
ine knowledge is possible only of what can be described; genuine descriptions
51
AVOWAL
or assertions involve the exercise of perceptual capacities, and the possibility
of observation (examination), justification and (dis-)confirmation.
Some readers of Philosophical Investigations §§243-315 have detected a pain­
ful emphasis on spontaneous expressions of pain. Even if (1) resembles an
expression like 'Ouch', this does not seem to be the case for psychological
terms which are not connected with a specific behavioural manifestation,
such as 'thinking'. But Wittgenstein's overall treatment does not in fact
suffer from a 'one-sided diet' of examples (PI §593). Moreover, Wittgenstein
acknowledges that any type-sentence can, in suitable contexts, be used non-
expressively, to make cool reports or explanations. Thus, a sentence like (1),
or an utterance like 'I am afraid', could be an expression, a report (e.g., to a
doctor) or an explanation (e.g., of one's trembling hands) (PI LT 187-9).
However, this concession invites the allegation that Wittgenstein is guilty of
a speech-act fallacy. The meaning of 'pain' must be the same whether it
occurs in avowals like (1), or in more complex cases, in which the sentence
does not serve to express a pain. The expressive role of some first-person
present tense psychological utterances seems due not to the meaning of the
words involved, but to the use to which they are put in the simple cases on
which Wittgenstein focuses. But Wittgenstein can reply: 'if "I'm afraid" is
not always something like a cry of complaint and yet sometimes is, then
why should it always be a description of a state of mind', as the inner/outer
dualism implies? (PI II 189; RPP I §633). He does not claim that psycholo­
gical terms are ambiguous, that, for example, 'pain' has a different meaning
1
in (1) from the one it has in (l ), but that (1) and (1') employ the term differ­
endy, are part of different linguistic techniques, and that the expressive use
of psychological terms is, in their first-person present tense application, the
standard one (RPP I §693; LW I §§874-5, 899).
Wittgenstein places excessive weight on the distinction between expressive
and descriptive uses. A single utterance can fulfil both functions: an utter­
ance of (2) can both state one's body weight and express remorse. More­
over, although 'I believe that p' is not a description, it is often a report
rather than a spontaneous manifestation; it may say what my long-held con­
victions are. Nevertheless, Wittgenstein is right to hold that psychological
reports are typically not based on inner observation or recognition of private
phenomena (RPP II §§176-7; LW I §51; PI §§274, 291-2; Z §434; see INNER/
OUTER). Moreover, they are parasitical upon genuine expressions of a pre-
linguistic kind (PI §§244, 290): unless certain forms of behaviour naturally
counted as manifestations of sensations, beliefs, emotions, etc., our mental
vocabulary would not have the meaning it does. This connection semanti-
cally characterizes, for example, sensation-terms. Although (2) may be used
to express remorse, that possibility depends on contingent assumptions
which are extrinsic to the meaning of 'weigh over 100 kg'. By contrast,
'pain' would no longer be the name of a sensation if avowals like (1) did not
52
AVOWAL
have a 'particular function in our life' which is analogous to that of natural
expressions of pain (LPE 301; LSD 35; RPP I §313; Z §§532-4). What dis­
tinguishes avowals from other utterances is the way they are linked to non-
linguistic forms of behaviour.
A final objection is that (1) is a basis for truth-functional operations like
conjunction, and can, moreover, function as a premise in a valid inference,
for example
(3) I am in pain; therefore someone is in pain.
Both points indicate that (1) is capable of being true or false, and is in that
sense descriptive. Furthermore, there is logical symmetry between avowals
and descriptions: (1), uttered by me now, says exacdy the same as (1') said
by you now, if I am N.N. And there are logical relations between the simple
cases on which Wittgenstein focuses, and complex cases. One defence of
Wittgenstein is that such functions and relations can involve uses of words
which are definitely not descriptive. From N.N.'s saying 'Off with his head'
we can infer something about N.N.'s state of mind (RPP I §463). However,
this is inadequate: what we draw an inference from here is not the state­
ment itself, but the fact that the speaker made it. By contrast, (1) appears in
inferences in its own right. (3) makes sense; 'Ouch; therefore someone is in
pain' does not. This is due to the fact that (1), unlike exclamations, is true
or false.
Fortunately, Wittgenstein grants that there are differences between first-
person psychological utterances and natural expressions (LPE 301, 318-20;
LSD 11; LW I §898). The former are articulate, that is, grammatically com­
posed of subject and predicate; can be used descriptively, and appear in
non-expressive contexts; allow of logical and tense transformations; and can
be true or false. But at the same time Wittgenstein insists that these simila­
rities to descriptions do not entail that avowals are straightforwardly descrip­
tive. As regards the symmetry between (1) and (1'), he would argue that
although their status is the same for the purposes of formal LOGIC, which is
concerned only with entailment, that is, transformations preserving truth-
value, it need not for that reason be the same for the purposes of philoso­
phical GRAMMAR. 'Being true' amounts to something different - has a differ­
ent grammar - in the case of avowals: their truth is guaranteed by
truthfulness (PI II 222), since they are not liable to mistake or error, only to
insincerity. Furthermore, although the sense of a proposition is not identical
with the method of its verification, 'whether and how a proposition can be
verified' is a contribution to its grammar (PI §353, II 224-5), which means
that the grammar of (1) differs from that of (1*). Wittgenstein's point is that
although avowals may be called descriptive, they lack conceptual connec­
tions which characterize ordinary descriptions (PI §§290-2; RPP I §572). He
53
AVOWAL
concludes that the inner/outer picture is mistaken to think that we 'read ofF
descriptions of our sensations, desires, thoughts, etc. from inner facts.
Finally, Wittgenstein recognized that with respect to first/third-person
asymmetry, psychological concepts form a spectrum of cases. At one end lie
sensations like pain, followed by intentions, thoughts, etc. Here there is no
such thing as being mistaken or finding out, and typically no room for
description. Somewhere in the middle lie emotions and states of mind with
genuine duration. They are typically avowed, but it is possible to find out
that I am in love or angry from my reactions. Similarly, I can describe the
course of my anxiety or fear as it waxes or wanes (PI §§585-8; RPP II
§§156, 722; LW I §43). But although here there is room for genuine self-
knowledge and error, which may sometimes rest on (mis-)perception or
(defective) observation, the problem is typically self-deception, a mistake of
the will, not of the intellect. At the other end of the spectrum are psycho-
pathological terms. I may (although I need not) be unqualified to decide
whether I am neurotic.
54
B
behaviour and behaviourism Modern philosophy has been dominated
by an INNER/OUTER dualism which distinguishes between the physical world
containing matter, energy and tangible objects, including human bodies, and
the private world of mental phenomena. Behaviourism is a twentieth-cen­
tury reaction to this position. It holds that attributing mental states, pro­
cesses or events to people really amounts to making statements about their
actual behaviour or dispositions to behave. Behaviourism comes in three
versions: metaphysical behaviourism denies that there are mental phenomena;
methodological behaviourism insists that psychologists should not invoke them
in explaining behaviour, since they are not intersubjectively accessible; logical
behaviourism claims that propositions* about the mental are semantically
equivalent to propositions about behavioural dispositions — thus
(1) Helga is sad
might be translated as
(1') Helga is speaking in a low monotone, and her head is drooping.
Wittgenstein has often been suspected of holding some version of beha­
viourism, and been placed alongside Ryle. His attitude to methodological
behaviourism is ambivalent. He claims that psychology, unlike philosophy,
has the task of investigating the causal mechanisms which link stimulus and
response. But this goes hand in hand with a 'hermeneutic' distinction
between understanding and explanation which implies that human action
cannot be made intelligible - seen as meaningful - through the CAUSAL
explanations of science (e.g. PLP ch. VI). UNDERSTANDING requires reference
to things which methodological behaviourism rejects - desires, beliefs,
moods, emotions, etc. His philosophy is also at odds with metaphysical
behaviourism. The early work presupposes that there is a language of
THOUGHT which consists of mental elements that can be studied by psychol­
ogy. Moreover, his first discussion of behaviourist ideas, those expressed in
Russell's account of iNTENnoNALTTY in The Analysis of Mind (chs III, XII), is
critical. He accuses them of mistaking the internal relations between an
expectation and its fulfilment, a symbol and its meaning, which are norma-
55
 BEHAVIOUR AND BEHAVIOURISM
tive, with the external relations between stimulus and response, which are a
matter of contingent fact.
Wittgenstein's relationship to logical behaviourism is more complex. He
never gave a behaviourist account of the first-person case. But a behaviour­
ist analysis of third-person psychological propositions is perhaps implicit in
the Tractatus (see BELIEF), at least Wittgenstein thought so in 1932 when he
accused Carnap of plagiarism in developing logical behaviourism under the
tide 'physicalism'. Such an account is explicit in the methodological SOLIP­
SISM of the transition period, which sharply distinguishes between genuine
'propositions', which can be verified by reference to primary experiences,
and third-person psychological propositions, which are mere 'hypotheses', to
be analysed in terms of behaviour. This position combines a 'no ownership'
analysis of first-person psychological propositions with a behaviourist analysis
of the third-person case (see I/SELF; PRIVACY). The official rationale for it was
provided by VERLFICATIONISM (WVC 49-50, 244; PR 88-95). If the meaning
of a proposition is the method of its verification, the meaning of third-
person psychological propositions like (1) is given by the behavioural evi­
dence we have for the mental phenomena (e.g., Helga's sadness). For we
cannot verify those phenomena by reference to the subject's private experi­
ences. Consequendy, to ascribe mental phenomena to others is to talk of
their behaviour. Wittgenstein also suggests that (1) has the same sense as (1'),
since both are confirmed by the same experiences. But even a verificationist
might resist this reductionist conclusion, on the grounds that there is possi­
ble evidence (not necessarily available) which sets apart (1) and (!') (e.g.,
Helga's laughing merrily when she is unobserved).
During the thirties, Wittgenstein became increasingly critical of behaviour­
ism, (a) He rejected the idea, to be found in Carnap's logical behaviourism,
that first-person psychological propositions can be analysed into propositions
about one's own behaviour, to be verified by self-observation. It does not
make sense to verify a proposition like 'I am sad' by observing one's own
posture and behaviour (PR 89-90; Z §539). Wittgenstein later claimed that
by and large such propositions are not descriptions at all, let alone descrip­
tions of behaviour, but AVOWALS, expressions of the mental. Such avowals
have a role similar to that of expressive behaviour, but they are not about
behaviour. To moan is not to say 'I moan', to cry out 'I am in pain' is not
to say 'I am manifesting pain-behaviour' (PI §244, II 179; LSD 11; LPE
296; RPP I §287).
(b) Against metaphysical behaviourism Wittgenstein stressed that it is
essential to the grammar of mental terms, even of sensation-words relatively
closely tied to behaviour, that someone can be in pain without manifesting
it, or that one can pretend to be in pain without being so. There cannot be
a 'greater difference' than that between pain-behaviour with and pain-beha­
viour without pain. At the same time, the PRIVATE LANGUAGE ARGUMENT
56
BEHAVIOUR AND BEHAVIOURISM
implies that the idea of pain as a private entity is a 'grammatical fiction' (PI
§§304-11) imposed on us by the AUGUSTINIAN PICTURE OF LANGUAGE which
suggests that words must refer to a 'something', which in the case of sensa­
tion-words is an inner something.
(c) While behaviourism rejects the Cartesian picture of the mind as a pri­
vate mental theatre, it accepts the attendant conception of the body as a
mere mechanism, and of human behaviour as 'colourless' physical move­
ments. Wittgenstein occasionally tended towards such a picture (PR ch. VI;
BB 51-2), but came to realize that it is flawed. The behavioural manifesta­
tions of most mental phenomena are extremely diverse. We can recognize
Helga's behaviour as expressing sadness only if ,we already approach it 'from
the point of view of sadness' (PR 89). This means that we do not, by and
large, infer psychologically relevant descriptions of human behaviour from
austere physical ones. For we often know the conclusions of such alleged
inferences without knowing their premises. It is easier to describe Helga as
'sad' or 'bored' than to describe her features or movements in physical
terms (RPP I §§1066-8, 1102; LW I §§766-7; Z §225).
(d) By a similar token, it is wrong to think that a HUMAN BEING is a body.
Rather, it requires a shift in perspective analogous to that involved in
ASPECT-PERCEPTION to view a human being as a physiological mechanism, and
human behaviour as mechanical movement (PI §420, II 178). For this
reason, Wittgenstein would not go along with Ryle's analysis of mental con­
cepts in terms of dispositions to behave. We can only ascribe mental con­
cepts to creatures with certain abilities. And, unlike dispositions, abilities are
(i) confined to sentient creatures and (ii) not actualized automatically given
certain conditions (one need not exercise an ability).
(e) When Wittgenstein speaks of the behavioural manifestations of the
mental, 'behaviour' includes not just facial expressions and gestures, but also
what people do and say, and the occasions for the use of mental terms.
These form a highly complex syndrome. What counts as a manifestation of
sadness on one occasion, may not on another (RPP I §§129, 314; Z §492).
The relationship between the mental and behaviour is much more compli­
cated than behaviourists suppose.
At the same time, Wittgenstein's later philosophy of psychology retains
points of contact with logical behaviourism. It rejects the dualist account of
the mental as inalienable and epistemically private. It accepts, albeit as an
empirical fact, that language-learning (and thereby the possession of a com­
plex mental life) is founded on brute 'training' (Abrichtung), rather than genu­
ine EXPLANATION, and presupposes natural patterns of behaviour and reaction,
to be activated by certain stimuli. And it claims that the ascription of psycho­
logical predicates to other people is logically connected with behaviour.
However, that logical connection is not one of logical equivalence between
propositions (namely psychological and behavioural ones). Rather, it takes
57
 BELIEF
two forms. First, it only makes sense to ascribe mental phenomena to crea­
tures who can manifest the mental in their behaviour. 'Only of a living
human being and what resembles (behaves like) a living human being can
one say: it has sensations; it sees; is blind; hears; is deaf; is conscious or
unconscious' (PI §281). Second, our mental terms would not mean what they
do if they were not bound up with behavioural criteria. The resulting posi­
tion undermines both behaviourism and dualism. Mental phenomena are
neither reducible to, nor totally separable from, their bodily and behavioural
expressions. The relationship between mental phenomena and their beha­
vioural manifestations is not a causal one to be discovered empirically,
through theory and induction, but a criterial one: it is part of the concepts of
particular mental phenomena that they have a characteristic manifestation in
behaviour (LPE 286; LSD 10). And it is part of mental concepts in general
that they have some such manifestation. We would have no use for these
expressions if they were not bound up with behavioural CRITERIA. If we came
across human beings who used a word which lacked any connection with
pain-behaviour and the circumstances in which we display it, we would not
translate it as 'pain'. The idea of super-spartans who are in constant agony
without showing it is as incoherent as describing as soulless human beings
who behave exacdy like us (IPP 281). 'The human body is the best picture
of the human soul' (PI II 178). We are inclined to think of mental episodes
as given, and of the expression as secondary, as mere symptoms through
which we may come to know the mind. But Wittgenstein makes out a strong
case for thinking that the intelligibility of mental terms presupposes the possi­
bility of behavioural manifestations. Ascribing THOUGHTS, for example, makes
sense only in cases where we have criteria for identifying thoughts, which
means that thoughts must be capable of being expressed.
belief Wittgenstein's earliest discussion of belief arises from his objections
to Russell's theories of judgement. Initially, Russell had held a dual-relation
theory, according to which a belief is a dual relation between something
mental - a subject or an act of belief - and a 'proposition', an objective
entity that exists whether or not it is believed. Tractatus 5.54f. dismisses this
theory as violating the extensionalist principle that when one proposition
occurs in another one, as according to the dual-relation theory p does in the
proposition '4 believes that p\ it can do so only as the basis of truth-func­
tional operations, which p does not in 'A believes that />' (for the truth of the
latter is not a function of that of the former) (see GENERAL PROPOSITIONAL
FORM). Both Wittgenstein and Russell also came to reject it for a less dog­
matic reason. In
(1) A believes /judges that p
what A believes is not an object, a fact. (1) does not presuppose that there is
58
BELIEF
something to be believed; it may be true even if no fact corresponds to p
(NL 95; Problems 72-3). In response to this problem, Russell developed his
multiple-relation theory of judgement (Essays ch. VII; 'Theory' 110): Othel­
lo's belief that Desdemona loves Cassio is not a dual relation between him
and a proposition, but a multiple relation between him and the constituents
of the proposition - Desdemona, love and Cassio. On this account, the
occurrence of the judgement does not entail that the relation of love obtains
between Desdemona and Cassio.
According to Wittgenstein, this ensures the possibility of false judgements
only at the unacceptable price of allowing nonsensical judgements. The cor­
rect analysis of (1) must 'show that it is impossible for a judgement to be a
piece of nonsense' (TLP 5.5422), '/»' must be a meaningful BIPOLAR proposi­
tion (RUL 6.13; NL 103). By splitting the proposition into its constituents,
Russell fails to guarantee the preservation of logical form between the con­
stituents of the judgement, and hence allows a judgement like 'the knife is
the square root of the fork' (similarly, Bradley complained that Russell
ignores the unity of judgement).
Tractatus 5.542 presents an analysis of belief intended to avoid both the
appearance that the proposition p here occurs in a non-truth-functional way;
and the possibility of judging nonsense. Wittgenstein's solution is to incorpo­
rate the propositional form in ascriptions*of belief. Thus, (1) is of the form
(2) '/>' says p.
Like (2), (1) correlates not a fact - p ~ and an object - the subject A - but
two facts, the depicted fact, p (assuming that p is a fact), and the thought-
constituting fact, '/>'. It does so through correlating their components,
namely the elements of thought with objects in reality. (1) means that in A
there is a mental fact which pictures the fact that p. Only composite things
with an articulate structure consisting of elements correlated with objects
can say or picture something. This implies that there is no unitary subject
'A', no soul-substance, but only a complex array of mental elements (TLP
5.5421; see SOLIPSISM).
This analysis guarantees the meaningfulness of what is judged by insisting
that it is not a complex of objects which can be combined in any old way,
but a FACT in which objects hang together subject to their combinatorial
possibilities. But it replaces Russell's inchoate notion of a relation between a
mind and the uncoordinated terms of judgement with the obscure idea that
'thinking the sense of/)' projects THOUGHT onto reality (TLP 3.11). More­
over, it is prima facie unclear how Tractatus 5.542 avoids the problem of
non-truth-functional occurrences. (2) can be understood in three different
ways. If what appears in quotation-marks is a description of 'accidental' fea­
tures of the propositional SIGN, (2) would always be false, since without a
METHOD OF PROJECTION signs cannot depict anything. Alternatively, (2) might
59
 BELIEF
express an external relation between two facts: the fact that the speaker
thinks or means such-and-such and the fact that p. In that case it is a bipo­
lar proposition, but its truth-value is determined not by that of '/>', but by
an empirical relationship between the fact that p and a mental fact. Finally,
the relation between the two facts might be internal, namely, if the descrip­
tion in quotation-marks includes a method of projection, that is, identifies 'p'
as precisely the proposition that says that p. But in that case (2) would be
necessarily true, and therefore its truth-value would again not be a function
of that of p. Moreover, by virtue of expressing an internal relation, (2) would
be a pseudo-proposition which tries to say what can only be shown by the
proposition p. None of these alternatives allows for '/>' to occur truth-func-
tionally in (2), or, consequendy, in (1). The last preserves the extensionality
thesis in so far as it does not violate the principle that propositions occur in
genuine propositions only as the bases of truth-functional operations, but does
so at the price of branding belief ascriptions as pseudo-propositions. Wais­
mann later suggested that the analysis should be confined to the first-person
case 'I believe that p' for which it was first developed (NM 119). But it is
hard to see how this avoids the aforementioned problems.
Wittgenstein's second discussion of belief starts out from Frege's and Rus­
sell's ideas about assertion (Notation §§2—3; Laws I §5; 'Function' 22; Corre­
spondence 79; Principles 35; Principia 8, 92; 'Theory' 107). Both distinguished in
an assertion the act of asserting from what is asserted, the proposition or
thought. One of their reasons was the need to distinguish the occurrence of
a proposition p when it is not asserted, as in 'p Z) q\ from its occurrence on
its own, when it is (the so-called 'Frege point'). For this purpose, Frege
introduced the assertion-sign 'p to express the act of judging something to
be true. Every line in his logical system has the form
(3) Yp
where '-/>' (involving the horizontal 'content-stroke') expresses the mere
thought without judging it to be true, while the vertical 'judgement-stroke'
signals the act of assertion which takes us from a thought to a truth-value. In
Frege's system, all inferences proceed from asserted propositions to asserted
propositions, and one can make inferences only from true propositions.
Having abandoned the idea that all judgements are of subject-predicate form,
he claimed that ' p (meaning 'it is a fact that') is the 'common predicate of all
judgements'. Russell took over the assertion-sign to add the force of 'it is true
that' to the unasserted proposition; he held that true propositions have the
quality of being asserted in a non-psychological, logical sense.
In 1911, Wittgenstein seems to have held that the only things which exist
are 'asserted' (i.e., true) propositions, which are facts. By the time of 'Notes
on Logic', however, he insisted that the assertion-sign is logically irrelevant
(NL 95-6; TLP 4.023, 4.063f, 4.442). It indicates merely the psychological
60
BELIEF
fact that the author holds the proposition to be true; it does not belong to
the proposition: (a) one can draw inferences from false propositions (Frege
and Russell ignored this, presumably because their axiomatic conception of
LOGIC focuses on proof, which requires true premises); (b) neither 'is true'
nor 'is a fact' is the 'verb' of propositions, the formal predicate they all have
in common; for what is asserted through these verbs must already have a
sense, that is, be a proposition.
As a result, Wittgenstein claims against Frege and Russell that logic is
exclusively concerned with the unasserted proposition, which shows how things
stand if it is true. However, this claim is inaccurate. For it is part of Wittgen­
stein's account that only the asserted proposition says something, namely that
things stand as the unasserted proposition shows them to stand. This is
required not just by the Frege point but also because the early Wittgenstein
seems to have accepted that an unasserted proposition can be common to
the assertion that p, the question of whether it is the case that p, the com­
mand to make it the case that p, etc. (TLP 4.022; NB 26.11.14; MS 109 249;
BT 149). Wittgenstein later returned to this idea (PI §22, 1 In; BT 208; RFM
116; Z §684; PLP 302-3). He ascribes to Frege the idea that that part of a
declarative sentence which expresses an 'assumption', that is, the thing that
is asserted, functions like a sentence-radical. The assumption, or thought, is the
descriptive content of what is asserted,••but can also be a component of
other, non-assertoric speech acts. It might be represented by '-p' in '?-/>' for
sentence-questions and '!-/>' for commands, as well as '\-p' for assertions. It
has been claimed that Philosophical Investigations uses this idea to accommodate
non-descriptive uses of language within the semantics of the Tractatus: the
PICTURE THEORY provides an adequate account of the sentence-radical, but
needs to be complemented by a theory of 'semantic mood' to account for
the uses of sentences in different language-games. In fact, however, the later
Wittgenstein rejects the idea that assertoric utterances can be analysed into
assumption plus assertion. He also rejects the idea that different speech acts
share a common propositional content, and that all propositions contain
descriptions. If sound, his arguments also undermine contemporary distinc­
tions between sense and force, and thereby threaten truth-conditional
semantics, which relies on the possibility of isolating in non-assertoric speech
acts a descriptive component (propositional content) capable of being true or
false (see AUGUSTINIAN PICTURE OF LANGUAGE). There are four points of attack.
(a) The Fregean theory imposes contradictory demands on that part of a
declarative sentence which is supposed to express the mere assumption or
thought. On the one hand, it must not be a complete sentence, since it must
lack assertoric force, as does the noun-phrase 'that />' in
(3') It is asserted that p.
On the other hand, it must be a complete sentence, since the assumption/
61
 BELIEF
thought is the sense of a sentence, not of a noun-phrase. Consequently,
there is no such thing as a sentence-radical, fit to express the sense of a
declarative sentence but unfit to express assertoric force. Assertion is not
something added to a proposition.
(b) One cannot characterize the concept of a proposition, what is true or
false, independentiy of that of assertion. But to this Frege could reply that
the non-assertoric occurrence of propositions in, for example, '/) D q' shows
that the concept of a proposition is at best linked to the possibility of asser­
tion.
(c) The assumption that p must be common to '?-/>' and '\~p'. According
to Wittgenstein, '?-/»' concerns the same assumption as '?-"/>', since both
questions can be answered by either '\-p' or '\-~p'. This implies, however,
that the assumption that p is the same as the assumption that ~p, which is
absurd.
(d) 'p does not signify a component of an assertion, or a mental activity
which gives the utterance its assertoric force. It merely serves as a punctua­
tion-mark which indicates the beginning of the sentence. And what gives an
utterance assertoric force is not an accompaniment, but the way it is used
by the speaker. But these observations are compatible with the Frege point,
the use of 'p to distinguish between the occurrences of '/>' in '\-p' and
*KO?)'.
The idea that assertion is a mental process which effects the transition from
a mere assumption to a declarative utterance is further attacked in the discus­
sions of 'Moore's paradox' from the 1940s (ML 10.44; PI II 190-2; RPP I
§§470-504; RPP IT §§277-83; Writings ch. 12). Moore had observed that while
we often do not believe something which is true, it is 'absurd' to say
(4) It is raining, but I don't believe it.
Wittgenstein rejected Moore's suggestion that this absurdity is of a psycholo­
gical nature. He claimed that utterances like (4) are senseless, and show
something about the logic of assertion. For one thing, they indicate a further
problem with the Fregean analysis. Moore's paradox shows that 'I believe
that it is raining' has a similar logical role to the simple assertion 'It is rain­
ing.' At the same time, the 'assumption' that it is raining is not the same as
the assumption that I believe this to be so, which concerns myself, not the
weather. If these two observations are expressed in accordance with the Fre­
gean analysis, we get:
(a) '|-/>' has a similar logical role to '\-R>p'
(b) '-/>' has a dissimilar role to '-Ib/>'.
While (a) implies that the assumption contained in '|-/>' is the same as that
in '(-lb/)', (b) implies that it is not. Moreover, (b) suggests that the assertion
'I believe that />' cannot be split up into an assumption and an expression of
62
BLPOLARITY
belief, because 'I believe' cannot be eliminated without altering the assump­
tion itself. Consequendy, the step from '-/>' to '\-p' cannot be one of adding
assertoric force to a common assumption.
A second implication of Moore's paradox is that belief is not a phenom­
enon which we observe in ourselves. If 'I believe that . . .' described in phe­
nomenal terms something about the speaker, whether about his brain, mind
or behaviour, (4) would not be paradoxical. For there could be no incon­
sistency between describing how things are with me (my mind/my brain)
and describing the weather. The role of 'I believe that />' is that of expres­
sing the belief that p. This is also a role of simply uttering '/>', which is why
there is an inconsistency between avowing '/)' and disavowing the belief that
p. I may report rather than AVOW my long-held convictions. But I do not
describe them, since such reports commit me to a claim, which no mere
description could (RPP I §§715-16; see INTENDING AND MEANING SOMETHING).
On such grounds, Wittgenstein attacks the neo-Humean position of James
and Russell, according to which belief is a feeling of approval towards a
proposition {Psychology II ch. XXI; Analysis 250-2). Although feelings may
accompany my beliefs, they are neither necessary nor sufficient. And
although 'to believe' is a static verb, it no more signifies a mental state such
as a feeling than a mental act or process. Belief is neither something one
does, nor something one undergoes or -ts in. Unlike genuine mental pro­
cesses or states, beliefs lack 'genuine duration' (PI §§571-94, II 193-229;
RPP I §§596, 710, 832-6; see PHILOSOPHICAL PSYCHOLOGY). Expressions of
belief are less determinate and characteristic than those of emotions, which
is why dispositional theories of belief fail. But the concept of belief is intern­
ally linked with what people would (sincerely) say they believe, and how they
would act in certain circumstances. (See also CERTAINTY.)
bipolarity According to the principle of bipolarity, every proposition must
be capable of being true, and capable of being false. This principle, which
Wittgenstein was the first to espouse, differs from the weaker principle of
bivalence, according to which every proposition is either true or false.
Represented symbolically (along lines Wittgenstein condoned in 'Notes dic­
tated to Moore' but later rejected), the principle of bivalence reads
(/>)(/) v ~p), while the principle of bipolarity comes out as (p)(0 p • O ~p).
From early on, Wittgenstein held that bipolarity is the essence of the propo­
sition (RUL 5.9.13; NL 94-9, 104; NM 113). The term derives from a
metaphor: a proposition, like a magnet, has two poles, a true one and a
false one. It is true if things are as it says they are, false if they are not. The
starting-point of this idea is Frege's view that names and propositions alike
have 'sense' and 'meaning', the meaning of a proposition being one or the
other of two 'logical objects', the True and the False. Initially, Wittgenstein
followed Frege in claiming that propositions have a MEANING, that is, stand
63
 BIPOLAR rTY
for something, just as names do. But he maintained that this meaning is not
a truth-value, but the FACT which corresponds to the proposition in reality.
The 'meaning' of p is identical with that of ~p, since the fact that makes it
true that p is the same fact that makes it false that ~p, and vice versa. But
the negation-sign reverses the sense of the proposition: capitalizing on the
ambiguity of the German Sinn (sense or direction), one can say that if it is a
fact that p, then the true pole of p points towards reality, and so does the
false pole of ~p. What p depicts is precisely what ~p depicts, only the latter
says that this is not how things stand.
Bipolarity marks a fundamental contrast between NAMES, which stand for
things, and PROPOSITIONS, which depict a possible state of affairs and can be
negated. This led the Tractatus to claim that only propositions have a sense
and only names a meaning. In order to understand a name one must know
its referent, but in order to understand a proposition one need not know
whether it is true or false. What we understand in the case of a proposition
is its 'sense', that is, both what would be the case if it were true and what
would be the case if it were false. Consequendy, the proposition is internally
related to its negation, somewhat as
^ is related to p ^ . To understand p is to understand
its negation (NL 97, 101; NB 14.11.14; TLP 3.144, 3.221).
The idea that it is essential for propositions to be bipolar contrasts with
Frege and Russell, and not just because they treated propositions as names
(of truth-values and complexes respectively). Frege went wrong, not only in
treating truth and falsity as objects which are named by some propositions,
namely those which do not suffer from truth-value gaps, but in ignoring that
a proposition is essentially connected with both truth-values. For him, there is
no closer connection between a true proposition and the False than between
that proposition and any other object (e.g., the number 7). To be sure, that
p is true entails that ~p is false; but Frege fails to realize that it is no coin­
cidence that negation operates in this way, but something which arises from
the very nature of the proposition p itself. Russell was closer to the Tractatus
in that he insisted on bivalence, and treated truth and falsity as properties
rather than objects. But he gave the impression that it is a contingent fact
that all propositions possess one of these properties. By contrast, Wittgen­
stein insisted on bipolarity rather than bivalence, and treated this as an
essential condition of a proposition's ability to represent reality (NL 104;
TLP 6.111-6.126, 6.2If.).
According to the principle of bipolarity, a propositional sign (Satzzeichen)
only has a sense if it determines a possibility which the world either satisfies
64
BIPOLARITY
or not. This has the astonishing consequence that logic, mathematics and
metaphysics do not consist of propositions. There can be no propositions
which are logically necessary, since they could not possibly be false, and
there would be no gap between understanding their sense and recognizing
their truth (cp. TLP 3.04f., 4.024). The truths of logic are TAUTOLOGIES, lim­
iting cases of meaningful empirical propositions, namely propositions with
zero sense. Metaphysical propositions are nonsensical. At best, they try to
say what can only be shown, the form of bipolar propositions. MATHEMA­
TICAL propositions are 'pseudo-propositions', they do not depict anything,
but are rules which license inferences between empirical propositions. The
TRUTH-TABLE presentation of propositions provides an ideal notation which
makes perspicuous the logical structure of all languages, because it shows
that propositions have essentially two poles (T and F). It also shows how the
necessary propositions of logic flow from this essential bipolarity, by display­
ing how in certain combinations the truth/falsity of elementary propositions
cancels out. This shows something about the structure of the world, namely
that it consists of mutually independent states of affairs (TLP 4.121, 6.12,
6.124; NM 108-11).
The logical positivists seized on bipolarity, and the ensuing treatment of
logical necessity, in order to exclude synthetic a priori truths. But Wittgen­
stein himself later rejected the principle of^bipolarity as part of a 'mythology
of symbolism' (PG 56; Z §211). In the Tractatus, propositions must be bipolar
because they depict states of affairs which either obtain or fail to obtain.
However, that facts either obtain or fail to obtain is not a metaphysical fea­
ture of reality, but merely part of what we call a fact or state of affairs.
Equally, TRUTH and falsehood belong to our concept of propositions, but this
is no metaphysical revelation, it merely means that we call propositions what
we also call either true or false (FW 55; PI §§136-7). Propositions are
indeed typically bipolar in that their truth excludes a possibility. But the
concept of a proposition is a FAMILY-RESEMBLANCE concept. There is no war­
rant for restricting the concept to descriptions of possible states of affairs.
Indeed, not even all empirical propositions are straightforwardly bipolar -
the WeltbiU propositions of On Certainty could not simply turn out to be false
(see CERTAINTY).
'The negation of nonsense is nonsense' (RAL 2 J.21). Wittgenstein later
relaxed this bipolar conception of NONSENSE by allowing that at least some
negations of nonsense, like 'Nothing can be red and green all over', are
GRAMMATICAL propositions. But the dogmatic principle soldiers on in some
parts of his later work: the claim that I cannot know that I am in pain,
because I could not be wrong, rests partly on the assumption that there is
no knowledge without the possibility of ignorance or error, and the sugges­
tion that T is not a referring expression on the assumption that referring
presupposes the possibility of referential failure. However, in these argu-
65
BIPOLARITY

merits Wittgenstein also pursues a more promising line. Rather than insist
dogmatically that the negation of a nonsense must itself be a nonsense, he
points out the difference between such propositions and propositions which
express a cognitive claim because they exclude possibilities that can be
described intelligibly, such as 'Nothing can be fatty and healthy' (see I/SELF;
PRIVACY).

calculus model Between 1929 and 1933, Wittgenstein compared speak­

ing a language to operating a logical or mathematical calculus (PR ch. XX;
BT 25, 142; PG 57, 63). The analogy serves different purposes.
(a) In speaking a language we operate, in thought, a complex system of
exact rules. The propositions of ordinary language can be definitely
analysed into elements of such a calculus (RLF; LWL 117).
(b) The meaning of a word is its place in the symbolism, it is determined
by rules which lay down its correct USE. Equally, to understand a sen­
tence is to see it as part of a system without which it would be dead.
'The role of a sentence in the calculus is its sense' (PG 130, see 59,
172; LWL 28, 37; BB 5, 42).
(c) Grammar is not a causal mechanism. The rules of a calculus specify
not what the probable result of employing a word will be, but what
sort of operation has been performed (PG 70).
(d) Speaking a language is an activity (PG 193; WVC 171-2), just as a
calculus is something we operate.
It has been maintained that Wittgenstein never abandoned the calculus
model. What is correct is that his later remarks continue to recycle material
from the early thirties which compares language with a calculus in order to
bring out points (b)-(c) (e.g. PI 14n, §§559, 565; MS130 214). But these
points are also expressed in the comparison of language with a game, in
particular that of chess. Although the term 'LANGUAGE-GAME' is first used as
equivalent to 'calculus' (PG 67), the fact that it replaced the latter by the
time of the Blue and Broum Books indicates a shift in Wittgenstein's concep­
tion of language. What remains is the idea that language is an activity gov­
erned by rules. What alters is Wittgenstein's conception of these rules: the
rules of GRAMMAR resemble those of a game like hide-and-seek more than
they do those of formal calculi. Finally, Philosophical Investigations explains the
calculus model as the view that 'if anyone utters a sentence and means or
understands it he is operating a calculus according to definite rules' (PI §81).
Wittgenstein states two things about this view, that he previously held it and
that it is mistaken.

66 67
 CALCULUS MODEL
Sometimes the calculus model is presented as the conception of language
which Wittgenstein held between the idea of logical syntax and that of
grammar. The analogy only emerges in the transition period, but the idea
of a language with precise rules goes back to Leibniz's plan for a character-
istka universalis. It was given a boost by Frege's and Russell's development of
logical calculi governed by a definite list of formation and transformation
rules. In their case the model was held to apply only to an 'ideal language'.
By contrast, the Investigations' characterization fits de Saussure's conception of
langue as an abstract system of rules which underlies parole, the use of ordin­
ary language on particular occasions, and the Tractatus's account of LOGICAL
SYNTAX as a system of rules for the meaningful combination of signs which
governs all symbolisms, including natural languages. These rules
are comprehensive and definite (TLP 5.4541): for any possible combina­
tion of signs, they determine unequivocally whether or not that combi­
nation makes sense, and, if so, what sense it makes, this sense being
itself 'determinate', namely a specific configuration of objects which
must obtain for the proposition to be true (TLP 3.23f; see MEANING);
constitute a highly complex system which is concealed by the school-
grammatical surface of language, and has to be discovered by LOGICAL
ANALYSIS;
govern human speech, although its speakers are not aware of them: 'Hu­
man beings have the ability to construct languages capable of expres­
sing every sense, without having any idea how each word has meaning
or what its meaning is - just as people speak without knowing how the
individual sounds are produced' (TLP 4.002, see 5.5562; RLF 171).
The Tractatus is committed to the view that speaking a language is operat­
ing a calculus of hidden rules. After his return to philosophy, Wittgenstein
claimed that this calculus does not reflect the essential nature of reality, but
is AUTONOMOUS. He also realized that ELEMENTARY PROPOSITIONS cannot be
logically independent. Propositions are compared with reality not individu­
ally, like pictures, but in groups, like the graduating marks of a ruler. Estab­
lishing that x is 3 m long ipso facto establishes that it is not 5 m long. Equally,
seeing that a point in the visual field is red implies ipso facto that it is neither
blue nor yellow nor green, etc. He concluded that propositions form 'pro­
position systems' (Satzsysteme), that is, sets of propositions such that their
members exclude each other not because of their trUth-functional complex­
ity but because of the concept-words occurring in them (WVC 63-4, 78-89;
PR ch. VIII; cp. TLP 2.15121). These systems of mutual exclusion are at
the same time LOGICAL SPACES of possibilities: 'black' is another mark on the
same ruler as 'red', but '5 m long' is not; the visual point could be black, yet
it could not be 5 m long (PR 75-7).
Accordingly, logical syntax is even more complicated than previously
68
CALCULUS MODEL
thought, which prompted Russell's complaint that Phiksophkal Remarks
would make 'mathematics and logic almost incredibly difficult'. By the
same token, analysis makes even more startling discoveries - for example,
that all propositions contain expressions for real numbers. Appearances not­
withstanding, natural languages are logico-syntactical systems. They consist
of formation and transformation rules and assignments of meanings to the
indefinables (which correspond to the 'axioms' of logical systems). These
joindy determine the sense of every well-formed sentence. Together with
the appropriate facts they also determine their truth-values unequivocally.
Wittgenstein quickly realized that the idea of proposition systems has only
a narrow range of application, namely to determinates (5 m long, red) of a
determinable (length, colour); and even there it ignores the fact that not all
determinates of a determinable share the same combinatorial possibilities (see
COLOUR). Gradually, he also came to attack the picture of language as a
system of precise and rigid rules. For one thing, linguistic rules are not
DETERMINATE in the sense of Frege and the Tractatus. They allow for border­
line cases, and do not invariably dictate for every conceivable circumstance
whether or not a combination of signs is NONSENSICAL. The same holds for
games: there are no rules for how high one throws tlfe ball before serving,
yet this does not mean that tennis cannot be played (PI §§68, 83; OC §139).
Indeed, the idea of an activity which is bounded by inexorable rules in all
of its aspects is absurd, since there are indefinitely many such aspects. More­
over, for any game there are coundess bizarre possibilities which cannot be
budgeted for in advance. The rules of tennis are none the worse for failure
to specify what happens if the ball is caught by a pelican flying by (PI §§80,
84-7; Z §440; PLP 76-80).
This insight transforms the Tractatus's insistence that 'all the propositions
of our everyday language, just as they stand, are in perfect logical order'
(TLP 5.5563). Ramsey had condemned this as a piece of 'scholasticism'.
Wittgenstein concurred by referring approvingly to Ramsey's remark that
'logic is a normative science' (Mathematics 269; PI §81). This reference
emphasizes the contrast between the clear, strict rules of logical calculi and
the fluctuating and vague rules of ordinary language (BT 248). It does not
acknowledge that ordinary language should, if possible, approach formal cal­
culi, as ideal-language philosophers like Frege, Russell and Carnap would
have it; Wittgenstein continues to reject the idea that ordinary language is
logically inferior to the formal languages of logic. Instead, it means that we
must not project the 'crystalline purity' of formal calculi onto ordinary lan­
guage by dogmatically insisting that complete order is hidden beneath a dis­
orderly surface (PI §§98-108). Formal calculi do not reveal the 'depth-
grammar' of language. Their only legitimate philosophical role is as objects
of comparison (PI §131; BB 28; MSI 16 80-2). They help us to achieve an
OVERVIEW of our grammar by way of similarity and contrast.
69
 CALCULUS MODEL
The most important element of truth in the calculus model concerns
points (c) and (d) above. Wittgenstein insists that a 'rule doesn't act at a dis­
tance. It acts only by being applied' (BB 13-14; PG 80-1). When I follow a
rule by Oing, rather than merely act in accordance with a rule, the rule is part
of my reason for acting as I do (see RULE-FOLLOWING). This means that rules
must somehow be involved in the process of explaining, justifying, applying
and understanding. For they have their normative status only in virtue of
being used as standards of correctness by us. Rules do not exist indepen-
dendy of the use speakers make of rule-formulations, their 'esse est appli-
cari'. In order to bring out this point Wittgenstein clarified rule-following by
reference to calculations in which the rule plays a visible part, such as con­
sulting rule-books or calculating according to a schema (WVC 168-71; PG
99-101; PLP 124-8).
At the same time, Wittgenstein had to acknowledge that most instances of
rule-following, including calculations, do not involve (overt or mental) con­
sultation of rule-formulations, just as competent chess-players rarely consult
the rules (WVC 153-4; LWL 48, 83, 101; PG 85-6, 153; PI §§54, 82-3;
RFM 414-22; PLP 129-35). A possible reaction is to insist that in such
cases the agent would explain or justify his <Kng by reference to the rule-for­
mulation (PI §§82-3). Rules have potential actuality. However, Wittgenstein
came to realize that even this does not hold inevitably. 'For not only do we
not think of the rules of usage . . . while using language, but when we are
asked to give such rules, in most cases we aren't able to do so. We are
unable clearly to circumscribe the concepts we use; not because we don't
know their real definition, but because there is no real "definition" to them'.
He concluded that regarding it as guided by definite or explicit rules is a
'one-sided way of looking at language' (BB 25; PG 68).
This gradual abandonment of the calculus model creates various tensions
in Wittgenstein's mature work. For one thing, why should one adopt this
one-sided perspective of cataloguing grammatical rules? Some passages sug­
gest that PHILOSOPHY sometimes 'makes up' or 'lays down' definite rules
where there are none, or accentuates aspects of linguistic use, namely for
the purpose of counteracting specific distortions of the concepts concerned
(AWL 47-8; BT 416; Z §467; RPP I §§51-2). Others insist that any filling
in of conceptual contours is itself a distortion (RPP I §§257, 648).
Moreover, even at a time when he was still comparing language to a cal­
culus, Wittgenstein castigated as 'hellish' Moore's idea that only logical ana­
lysis shows us what, if anything, we mean by our propositions (WVC 129—
30). More generally, he rejected the idea which united Frege, Russell and
the Tractatus, that analysis can make 'deep' or 'unheard of discoveries; there
are, he insisted, no surprises in grammar (WVC 77; LWL 16-17; BT 418—
19, 435-6; PG 114-15, 210; MS109 212; MSI 16 80-2). 'What is hidden, is
of no interest to us' (PI §§126-8). This is justified if grammatical rules are
70
CALCULUS MODEL
evident in the explanations speakers would give when asked. However, that
fails to be the case, not just for the FAMILY-RESEMBLANCE concepts at issue in
Blue and Brown Books 25, but also, for example, for the difference between
'almost' and 'nearly', the use of the definite article or of the subjunctive, and
the sequence of tenses in conditionals. Wittgenstein would accept examples
as adequate EXPLANATIONS. But in these cases, even they may not be forth­
coming. Consequendy, in so far as language is governed by GRAMMATICAL
rules, these are not simply open to view, but. need to be made explicit (in
line with Ryle's distinction between 'knowing how' and 'knowing that').
Wittgenstein is right that this is not a matter of gathering new information -
as competent speakers we have all the information we need. But it is a
matter of elicitation and reflection, and may involve trial and error.
Ironically, questioning one side of the calculus model, the idea that we
constandy consult rule-formulations, leads one to another, the idea of dis­
coveries. But Wittgenstein's attack retains critical potency against positions
which combine both ideas. This holds of Wittgenstein's own early position,
and for some contemporary theories of meaning for natural languages in
philosophy (Davidson, Dummett) and linguistics (Chomsky). These theories
are committed to the view that we have tacit knowledge of a complex
system of formation and derivation rules which is hidden below the surface
of language as presented by school-grammar. Meaning and understanding a
word consist in operating this calculus; but since we are not aware of such
calculations, they must be subconscious, and occur at high speed.
Against this, Wittgenstein shows that UNDERSTANDING does not require any
such calculations. The causes of my speaking and understanding include
high-speed neural processes unknown to me, but this does not hold of my
reasons for applying or understanding words in a certain way. Although the
rules reconstructed by philosophical grammar may not play a role in our
practice of applying and even explaining words, it is assumed that speakers
are capable of recognizing certain formulations as expressions of the rules
they are following. What is important is that they should recognize these
formulations not just as accurate descriptions of patterns of linguistic beha­
viour, but also as expressing standards by which they distinguish correct and
incorrect employments of words. For example, speakers who may be incap­
able of explaining the terms 'automatically' and 'inadvertendy' will recognize
that a certain form of behaviour can satisfy the latter term without satisfying
the former.
Even this potentiality is absent in the case of the rules invoked by theories
of meaning. Indeed, the arcane apparatus of the latter is unintelligible to
many competent speakers. The inference from 'She kissed him in the
garden' to 'She kissed him' is recognized by people who are incapable of
even learning the quantificational rules through which theories of meaning
explain its validity. This means that such rules are in no sense standards of
71
 CAUSATION
the correct use of our words (RFM 414-22; MS129 79). The line between
rule-following and acting in accordance with a rule has been blurred.
Wittgenstein is right to insist that rules cannot be hidden in the sense that
we are denied access to them, or transcendent, incapable of playing a role
in our practice. However, he fails to show that we cannot make discoveries
of some kind. Instead, he makes out a case for thinking that these will
reveal language to be structured not by logical calculi, but by the diverse
and complex patterns and subde nuances highlighted by ordinary language
philosophy.
causation Wittgenstein's early account of causation has both a negative
and a positive side. Negatively, he follows Hume in rejecting the idea of
causal necessity. There is only one kind of necessity, namely logical neces­
sity; 'outside logic everything is accidental.' This means that there is 'no
causal nexus' to justify an inference from the existence of one situation
(Sacklage) to that of another. Hence, too, there is no 'compulsion' that one
thing should happen because another has happened, and we cannot know
future events (TLP 5.135-5.1362, 6.3, 6.36311-6.372; PT 5.0442f; NB
15.10.16; see INDUCTION). Positively, Wittgenstein explains the role of causa­
tion in SCIENCE through a Neo-Kantian account of natural or causal laws.
Like other so-called 'fundamental' laws of science, the 'law of causality',
according to which every event has a cause, is not a law, but 'the form of a
law'. This means that it is neither a law of logic, nor an empirical general­
ization, nor a synthetic a priori proposition (see INDUCTION). Indeed, it is not
a proposition at all, since it tries to say what can only be shown. What it
indicates is a certain 'form of description' which is crucial to scientific theo­
rizing (TLP 6.321f.). Descriptions which connect events in a non-lawlike
manner are excluded from science. To characterize something as an event is
to imply that it is explicable by reference to some (often unknown) causal
law. Causation itself is a formal concept. It characterizes not reality, but the
'network' of an optional FORM OF REPRESENTING reality, such as Newtonian
mechanics (TLP 6.33-6.341, 6.36f., 6.362).
Wittgenstein's later thoughts on causation, assembled mainly in 'Cause
and Effect: Intuitive Awareness' (see also LC 13-15; BT 406-7), move away
from the empiricism of the negative account, while developing the con­
ventionalist themes of the positive account. He continues to hold the
Humean view that causal relations are external, that is, obtain between logi­
cally independent events (see PI §220; Z §296), and to elucidate causation
by reference to causal explanations. But he now focuses on the way we
establish causal connections in everyday life, and the results challenge cru­
cial aspects of the Humean position.
Firsdy, he rejects a uniform nomological account of causation. There is
an irreducible variety of 'prototypes' of causal connections: (a) impact (colli-
72
CAUSATION
sion of billiard balls); (b) traction (pulling a string); (c) mechanisms Like
clocks, which combine (a) and (b); (d) human reactions to sensations or emo­
tions (being hit on the head or frightened by someone's facial expression); (e)
statements which are based on observing regular successions of events. Since
Wittgenstein stresses both the variety of cases and the fact that we use the
same word, he arguably regards 'cause' as.a FAMILY-RESEMBLANCE concept. He
denies not only that the Humean paradigm (e) is the only prototype of cau­
sation, but also that it is the fundamental one. The 'cause-effect language-
game' of everyday life is rooted not in observation or experimentation, but
in a practice, which in turn is based on certain primitive reactions. For
example, we react to a painful blow by pointing to someone and saying 'He
did it' (CE 409-10, 416-17, 420, 433).
Secondly, according to Hume we can never direcdy observe a causal con­
nection, but only a succession of events; consequendy our causal statements
must be based on observing a regular sequence of parallel events and are
always provisional, subject to refutation by subsequent observations. Witt­
genstein follows Russell in holding that there are causal relations which we
know immediately, while rejecting the idea that this is based on intuition
(CE 409, 431; LC 22). Recognizing the most basic forms of causation, espe­
cially those involving direct physical contact, (a)-(d), does not depend on
observing constant regularities or on experiments; we direcdy observe one
thing acting upon another, and know the cause immediately, though not
infallibly. Both immediate and non-immediate connections are paradigmatic
cases of what we call a causal nexus, and constitutive of the idea of causa­
tion. While rejecting Hume's empiricism, Wittgenstein makes the claim that
the principle of causality, 'Every event must have a cause', is not a synthetic
a priori truth, as Kant thought, but a disguised rule of GRAMMAR (AWL 16).
If this means that our grammar simply rules out as nonsensical the expres­
sion 'uncaused event', it is wrong. But one might argue that it is a norm of
representation of classical mechanics that it always makes sense to hok for
the cause of an event, even if no plausible candidate is in sight.
Wittgenstein also challenges a more general dogma which unites empiri­
cists and rationalists, namely that all causes must be necessitating: whenever
an effect occurs in one case but not in an apparendy similar case, there
must be relevant further differences. Wittgenstein, by contrast, denies that
in the case of two apparendy identical plant seeds which produce different
kinds of plants there must be a difference in the seeds underlying these dif­
ferent dispositions. The insistence that there must, is not based on an
insight into the actual nature of things, but amounts to adhering to a norm
of representation - instead, we could treat the origin of the seeds, irrespec­
tive of their physical structure, not just as the basis for a prediction ('Seeds
from a type-A plant will produce type-A plants'), but also as a genuine
explanation, that is, add ' . . . because they are from type-A plants'. He even
73
 CAUSATION
suggests that it would be better to abandon this norm (Z §§608-10; CE
410-11, 4 3 3 ^ ) .
One might concede that there is nothing unintelligible in supposing that
there is no structural difference between the seeds, but insist that looking for
such a difference come what may is a Kantian regulative principle, which is
constitutive of scientific investigation and perhaps of rational thought. Here
it is important to separate several issues. Wittgenstein is right to reject the
idea that all dispositional qualities must be explicable in terms of structural
properties of the objects manifesting them. For this could not apply to the
ultimate constituents of matter, which, by definition, have no components
and hence no structural properties. Wittgenstein is also right in claiming
that the idea of necessitating causes is an optional norm of representation.
Indeed, there are areas of science that operate with non-necessitating causes,
notably quantum-mechanics. However, as Wittgenstein himself acknowl­
edges, his treatment of the seed example upsets our conceptions of causality,
since it urges us to accept explanations by reference not just to non-necessi­
tating causes, but also to phenomenal properties (concerning the origin of
the seeds). Accepting such explanations is on a par with accepting astro­
logical explanations backed by statistical evidence. It amounts to abandoning
a norm of representation - 'Causal explanations must ultimately be struc­
tural, not phenomenal' - which may not be justifiable by reference to an
'essence of reality' (see AUTONOMY OF LANGUAGE), but which has been definitive
of scientific explanation since the seventeenth century.
Wittgenstein's idea of non-necessitating causes has been taken up by
Anscombe. The claim that there is an irreducible variety of types of causa­
tion and that the notion of a cause is rooted in action rather than observa­
tion is reminiscent of Collingwood's idea of a cause as a 'lever', a condition
under the control of human agents by means of which they can bring about
or prevent certain other conditions. Von Wright has defended the stronger
claim, suggested by Wittgenstein, that the interventionist notion of 'cause' is
not only genetically but also logically prior to the one based on observation,
since it alone affords the means of distinguishing between mere con­
comitance and a genuine causal connection.
A central feature of Wittgenstein's later reflections is that causal explana­
tion is only one way of answering the question 'Why?', and that reasons
must be distinguished from causes. He blames the ubiquitous temptation of
assimilating the two on the fact that reasons, like Humean causes, are gen­
eral, and on the impression that in the first-person case we are aware of our
reasons as causes 'seen from the inside' (BB 15; see PG 228; PI §378; PLP
119-22). He makes a few points to distinguish the reasons for believing that
p or for Oing from the causes, often in the context of criticizing Freud's
view of psychoanalytic explanations as causal (though unfortunately without
developing them at any length).
74
CAUSATION
(a) The concept of a reason is that of a step in reasoning, which is a tran­
sition from one assertion or thought to another: 'Giving a reason is like
giving a calculation by which you have arrived at a certain result' (BB 14—
15; RFM 39; AWL 4-5; LC 21-2; PI §§489-90). This need not mean that I
actually passed through a certain process, but includes justifications ex post
acta, which invoke steps I could have taken. The difference between asking
for the cause and asking for the reason is like that between asking 'What
mechanism has taken you from A to B?' and asking 'By what route did you
come from A to B?' Reasons, unlike causes, play a justificatory role. More­
over, in some cases the relationship between a reason and what it is a
reason for is INTERNAL, that is, (partly) constitutive of the relata, like the rela­
tionship between the premises and the conclusion of a LOGICAL INFERENCE
(deductive argument), or between a rule and its correct application (see RULE-
FOLLOWING).
(b) We must typically know our reasons, and the criterion for what a per­
son's reasons are is what that person sincerely avows them to be. (The Freu­
dian conception of 'unconscious reasons' modifies the concept of a reason,
but still insists that ascribing unconscious reasons is subject to the patient's
consent.) Unlike causes, reasons for one's behaviour are not discovered by
one on the basis of evidence (AWL 5, 28, 37-40; BB 57-8; LC 18, 23-5;
PG 101; PI §§475, 487-8; LPP 23).
(c) While chains of causes go on indefinitely, reasons come to an end.
Even where there are chains of reasons, these peter out. But this does not
open the door to SCEPTICISM; it is part and parcel of the concept of justifica­
tion (BB 14; PI §§217, 485).
Wittgenstein's distinction between reasons and causes is at odds with a
causal conception of the mind, according to which mental phenomena are
the inner causes of outward behaviour. Part of this picture is a causal con­
ception of intentional action, for which human behaviour is explained by
reference to efficient causes - acts or events which take place either in a pri­
vate mental realm (the soul) or, more plausibly, in the brain. Wittgenstein,
by contrast, holds that intentional behaviour is explained teleologically, by
reference to an agent's reasons (beliefs, intentions, wants). Unlike efficient
causes, reasons do not necessitate action: if the agent could do no other he
would not act intentionally. This view stands in the tradition of the herme-
neutical distinction between the explanation (Erkldreri) of the natural sciences
and the understanding (Versteken) of the social sciences. It was developed by
Anscombe, who, unlike Wittgenstein, claimed explicidy that the link between
an action and the reason for action is always internal, and hence not causal;
and by Winch, who linked it to the methodology of the social sciences.
Their position has been forcefully criticized by Davidson. Starting out from
the Wittgensteinian idea that logical relations are de dicto, that is, due to the
way we describe things, he argues that reason and action may be logically
75
 CERTAINTY
related under some descriptions, but not others, which leaves open the pos­
sibility that, besides being linked internally, they are events related by causal
laws. Moreover, what else could explain the difference between something's
being a reason for an action, and something's being the reason why it was
performed, if not that only the latter was causally efficacious in bringing
about the action? Davidson concludes that although we explain action by
reference to reasons (beliefs and desires) these are causes and are identical
with neurophysiological phenomena. According to Wittgenstein, on the
other hand, the correlation between mental and neurophysiological phenom­
ena is merely contingent; it is not logically necessary that the mental life has
causal roots (see INNER/OUTER). He also denied that beliefs and desires are
mental states with genuine duration, which implies that they cannot be iden­
tical with neural states (see PHILOSOPHICAL PSYCHOLOGY).
Wittgenstein intimates a way of resisting Davidson's 'what else?' argument
(PI §487; AWL 5; LC 22-3). We distinguish the reason for Oing from other
reasons by reference not to the presence of a causal connection, but to the
context of the action, notably to what reasons have previously weighed with
the agent in similar circumstances. Indeed, there need not be a pre-estab­
lished connection between the action and the reason. Often it is only the
agent's sincere avowal which determines why he did it, although sometimes
the context provides grounds for rejecting such avowals as based on self-
deception (contextualist elements are also central to Wittgenstein's attacks
on causal conceptions of the WILL). However, even if correct, this argument
does not rule out that some mental concepts are causal. When I say that I
clenched my teeth because of the stabbing pain in my neck, I do not give
reasons for the clenching, I give a causal explanation.
certainty Many of Wittgenstein's discussions have implications for episte-
mology. But it was only during the last year and a half of his life that he
tackled the topic in a direct and sustained way. The resulting notes have
been published as On Certainty. They were never polished or revised, let
alone completed, and hence contain numerous hesitations, occasional incon­
sistencies and much inconclusiveness. But they also possess a thematic unity
absent from most of Wittgenstein's later work.
On Certainty's inspiration was Moore's defence of common sense. Moore
maintained that there are empirical truths which one can know with cer­
tainty, for example, that one is a human being, that the object one is point­
ing to is one's hand, and that the earth has existed for many years. Moore
thought that these 'common sense' propositions are founded on evidence,
although we often cannot tell what it is, and that they entail that there is a
mind-independent world, and thereby refute scepticism. Wittgenstein
thought that Moore had drawn attention to an important class of proposi­
tions. He granted that one can be certain of these truisms, but denied that
76
CERTAINTY
one knows them. He granted that 'I know that />' where p is a Moore-type
proposition can have an everyday use in exceptional circumstances (OC
§§23, 252-62, 347-50, 387, 412, 423, 433, 526, 596, 622; Writings chs 3, 9 -
10). But this is not Moore's point, since it does not provide a reply to scepti­
cism. Moore's use ignores that in 'normal linguistic exchange' (OC §260, see
§§58, 243; PI II 221) we reserve 'I know' for cases in which
(a) it also makes sense to speak of believing or making certain;
(b) there is an answer to the question 'How do you know?';
(c) one is prepared to give compelling reasons -for one's claim.
Requirement (b) does not mean that we must actually be able to answer
the question, only that there is an answer in principle. Nor does it commit
Wittgenstein to the view that all knowledge rests on evidence. The question
'How do you know?' can equally be answered by specifying the perceptual
faculty through which one has informed oneself of something. Wittgenstein
occasionally suggests that we can talk of knowledge only where, (i) there is a
logical possibility of being mistaken or ignorant, and (ii) that possibility has
been ruled out by the application of 'clear rules of evidence'. But he also
states that 'I know how it is = I can say how it is, and it is as I say it is',
which implies that I can know in the abs*ence of these conditions (LW II 49,
58; OC §§243, 250, 483-4, 564, 574-6).
Another of Wittgenstein's points is that Moore's assurance that he knows
that he has two hands does not ensure that he does know, for while avowals,
like 'I believe' or 'I am certain', guarantee belief or certainty, 'I know' as an
expression of a conviction of knowledge does not guarantee knowledge, only
that one thinks one knows (LW II 89; OC §§12-15, 21, 137, 180, 489). The
certainty involved here is what Wittgenstein calls 'subjective certainty', a
feeling of unshakable conviction. But he also suggests that objective cer­
tainty, which is not a mental state, but signifies the inconceivability of doubt
or of one's being mistaken, belongs to a different category from knowledge
(OC §§54-6, 193-4, 308; LW LI 88). Wittgenstein does not substantiate this
claim, but may nevertheless have a point against Moore. The use of 'I
know' outside its normal context invites the sceptical question of how this
has been settled. Moore seems to make a 'presumptuous' and 'uncondi­
tional' claim, that nothing could prove him to be wrong. But 'I know' seems
not to tolerate such a 'metaphysical emphasis' (OC §§21, 251, 425, 481-2,
553-4). Like Austin, Wittgenstein suggests that empirical knowledge claims
are defeasible: even if they are well justified, there can be no metaphysical
guarantee against their turning out to be wrong.
On the other hand, this does not license scepticism. Doubt requires
grounds. But the mere imaginability of not-/) is no ground for doubting that
p (OC §§4, 122, 323, 458, 519). This is obvious if imaginability signifies the
77
 CERTAINTY
logical possibility of not-/>, which obtains by definition f6r contingent propo­
sitions. But the Cartesian sceptic will claim that our being deceived by a
malignant demon is an epistemic possibility, that is, consistent with what we
know. But that suggestion in turn requires a rationale, and the only ratio­
nale for the malignant demon hypothesis is its logical possibility. The Carte­
sian sceptic validly infers from the possibility that I shall turn out to be
wrong the possibility that I do not know. But he is wrong to infer from the
latter that I do not know (O ~p D O ~Ka/ but not O ~Ka/> D ~Ka/>). In
cases in which my well-supported claim turns out to be right, I did know.
The possibility of 'I thought I knew' does not count against the possibility of
'I know' (OC §12).
Wittgenstein sometimes concedes Moore's use of 'I know' and con­
centrates on the crucial point, namely the contrast between these uses and
ordinary empirical knowledge claims (OC §§288, 397, 520, 552). Moore's
propositions play a 'peculiar logical role in the system of our empirical pro­
positions'. They constitute the 'scaffolding' of our thoughts, the 'foundations'
of our language-games, the 'hinges' on which our questions and doubts
turn, our 'world-picture', 'the inherited background against which [we] dis­
tinguish between true and false' (OC §§94-5, 136, 211, 308, 341-3, 401-3,
614, 655). Hinge propositions are empirical in that their negation makes
sense. But the possibility of their being false is restricted by the fact that our
whole system of beliefs depends on our ready acceptance of them. Doubt
concerning them is infectious, and does not fit with anything else we believe.
Among the world-picture or hinge propositions listed by Moore and Witt­
genstein one can distinguish four types (OC §§4, 118, 207, 281-4, 291-3,
327, 555-8, 567, 599, 618). The first are trans-historical: they stand fast for
any sane person - for example, 'The earth has existed for a long time' and
'Cats don't grow on trees.' The second change with time: they were origin­
ally discovered and supported by evidence, but, once established, occupy a
pivotal role in relation to others, such as that there is a brain in the human
skull or that water boils at 100° C. In addition to these impersonal hinge
propositions there are two types of personal cases: generally applicable pro­
positions about which each person is certain for himself, such as T have two
hands' and 'My name is N.N.'; and person-specific propositions which are
part of my subjective world-picture, for example that I have spent most of
my life in Germany.
Wittgenstein makes a variety of claims about hinge propositions, (a) They
are (mutatis mutandis in the last case) certain not just for individuals, but for
everybody, unlike for example the claim that in a certain part of England
there is a village called so-and-so (OC §§100-3, 462).
(b) I might, in special circumstances, be wrong about hinge propositions,
but that would mean that I am deranged rather than merely mistaken. Poli­
ticians on a campaign tour are often mistaken about where they are, but if
78
CERTAINTY
they did not know where they usually live, that would be an aberration. I
might be confused about my own name, but if this happened to a majority,
the language-game with proper names could not be played (OC §§71-5,
156, 303-5, 579, 628).
(c) Hinge propositions of the trans-historical kind are not based on investi­
gation and are not supported by evidence (OC §§103, 138), because there
are no more fundamental propositions on the basis of which they could be
believed. There is evidence for these propositions in the sense that they
could be defended by offering certain considerations. But these are not my
reasons for holding the belief, because they are not better known to me
than the conclusion, although they might do that for people with a different
set of beliefs. We have geological and evolutionary evidence, for example,
for
(1) The world has existed for a billion years
but not for
(2) The world has existed for a hundred years.
Although (1) entails (2), it does not support it. For me evidence in its favour
presupposes (2): although it is not direcdy derived from it, that evidence,
along with the whole discourse of geological evidence, would collapse with­
out (2).
(d) Wittgenstein also suggests that the sense of hinge propositions like (2) is
less clear than that of empirical propositions like (1) because it is unclear
with 'what ideas and observations' they belong (PI II 221-2). However,
unlike 'I know that I am in pain', to which he links it (2), many hinge pro­
positions exemplify Gricean conversational implicatures, since their negation
is not nonsensical. Although in normal conditions hinge propositions are too
obvious to be informative, and are not held by us on the basis of evidence,
their conventional sense specifies what evidence we could use. A king who
believed that the world started with him would have to be converted to our
world-view (OC §§92, 422), but it is clear what we would use to effect the
conversion (photographs, written testimony, etc.).
(e) By far Wittgenstein's most important claim about hinge propositions is
that they can be neither justified nor doubted, since their certainty is pre­
supposed in all judging (OC §§308, 494, 614). One of his points is that
doubt cannot stand at the beginning of the LANGUAGE-GAME. If a child were
immediately to doubt what it is taught, it could not learn certain language-
games. But the point is not just either genetic or pragmatic — due to our
human condition we must begin with non-doubting. The suspicious pupil is
not displaying admirable caution, but simply fails to participate in our epis­
temic practice, and hence to raise a genuine doubt. Doubt only makes sense
within a language-game. By extending doubt to the hinges on which the
79
 CERTAINTY
language-game turns, the SCEPTIC is sawing off the branch on which he is sit­
ting. Doubt presupposes not only the possibility of certainty, but that many
things are certain. Our language-games can only be played against a rela­
tively permanent backdrop of certainties (OC §§115, 150, 283, 472-7).
One line of argument here is holistic: we must hold certain tilings certain
in order to question others. This leaves open the possibility that individual
hinge propositions should turn out to be wrong. Indeed, commentators have
felt that Wittgenstein's inclusion of 'I know that I have never been to the
moon' (OC §§106-11, 286, 662-7) shows that he overestimated how central
hinge propositions are to our web of beliefs. But as we have seen, Wittgen­
stein acknowledges that some hinge propositions may loose their status.
Moreover, to suppose now that Wittgenstein in 1951 might have been to
the moon is not the same as to suppose in 1951 that travelling to the moon
might become possible. For Wittgenstein, it raised unmanageable questions
about how he should have escaped the earth's gravitational field, etc.
Whether the revision of a hinge proposition will lead to a collapse of our
web of beliefs partly depends on whether we are dealing with a change of
natural processes, or a discovery. Certain scientific discoveries would not pre­
vent us from engaging in most of our language-games, but would merely
change the discipline concerned. But what 'if something really unheard-ofhap­
pened', for example, that now cows stood on their heads and laughed and
spoke (OC §§512-18)? Wittgenstein intimates (with Austin) that this would
not so much show that I did not know that this is a cow as show that what
used to be a cow has changed into something else. Unheard-of events do
not so much falsify our claims as lead to a breakdown of our concepts. In
some cases this change would be restricted to particular concepts. But if nat­
ural regularity broke down, our practice of making knowledge claims might
lose its applicability altogether.
Wittgenstein grants the possibility of unheard-of events, or of my suddenly
being contradicted from all sides. Some have detected in this a kind of
meta-scepticism. This seems borne out by Wittgenstein's idea that we cannot
know the truth of hinge propositions, or at any rate would have to qualify
any claim to such knowledge by 'so far as one can know these things', and
by his claim that it 'is always by favour of Nature that one knows some­
thing' (OC §§420, 503-5, 623). But his point is merely that it is a contingent
fact that nature is such that we can operate with certain concepts like
knowledge. He even leaves open whether we could continue our language-
games even if these FRAMEWORK conditions changed (OC §§516, 619). The
mere logical possibility of unheard-of events does not license Humean Angst
that chaos might break out at any moment, since unheard-of events are
excluded by natural necessity (although at the ultimate, micro-physical level
what is naturally necessary is a matter of brute fact - see CAUSATION).
Wittgenstein occasionally spoke of hinge propositions as providing the
80
COLOUR
foundations of our rational thought (OC §§ 162-7, 401-2). But these founda­
tions do not serve as a basis for other beliefs in the sense in which axioms
underpin theorems. We don't deduce other truths from them, but rely on
them as a 'background' for our rational arguing. The 'foundation-walls are
carried by the whole house', that is, they owe their special status to the fact
that they underlie the linguistic institution of argument (OC §§246-8, see
§§94, 153, 204). Indeed, the ultimate foundations of our knowledge are not
beliefs, but forms of behaviour.
According to Wittgenstein, the certainty of a belief consists in its role
within our framework of beliefs. A belief is certain if it can be appealed to
in order to justify other beliefs but does not itself stand in need of justifica­
tion. Descartes would protest that this does not answer the sceptic, since the
latter doubts whether these beliefs should play that role. However, that chal­
lenge presupposes that these practices must_ reflect the essence of reality,
which runs counter to the AUTONOMY O F LANGUAGE.
On Certainty's most important achievement was to provide the cue for an
epistemology socialized (which the sociology of knowledge claims to have
taken up). Neither the knowledge of a culture, nor even that of any of its
members, can be derived from the perceptual experiences of an individual.
The accumulated knowledge of a culture is a collective achievement - an
idea shared by Hegelians, Marxists and pragmatists. None of us can survey,
let alone master, that totality (OC §§161, 288-98). Learning is based on
accepting the authority of a community, and even adults have to take many
things on trust (OC §§170, 374-8, 509; Z §§413-16). But this does not deny
the possibility of critical thought. By accepting many things, we can partici­
pate in epistemic activities which allow us to rectify some of our beliefs,
occasionally even parts of our world-view (OC §§161-2). Unlike Quine's
epistemology naturalized, Wittgenstein's epistemology socialized makes this
point without reducing belief-formation behaviouristically to a matter of
stimulus and response.
colour This provides an illustration of the atomistic ontology of the Trac­
tatus. Wittgenstein makes three points:
(a) There are internal relations between colours, relations which cannot
fail to hold between them, for example that white is lighter than black
(TLP 4.123).
(b) 'Being coloured' (along with space and time) is a 'form of objects'.
Every 'visual object' (visible object) is in a 'colour-space', that is, it
must have some colour (just as every object must have some spatio-
temporal location); this is one of its essential 'internal properties' (TLP
2.0131, 2.0251; PT 2.025If.; see LOGICAL FORM).
81
 COLOUR
(c) Ascriptions of different colours to a point in the visual field are incon­
sistent.
(1) A is red
necessarily excludes
(2) A is green (blue, yellow, etc.).
Such 'colour-exclusion' - (c) - is an apparent counter-example to the
Tractatus's claim that all necessity is LOGICAL, a consequence of the truth-
functional complexity of molecular propositions. Wittgenstein tries to deal
with this difficulty by showing that (1) and (2) can be analysed as logical
products which 'contradict' each other (e.g., (1) as 'p.q.r'; (2) as 's.t.~r').
He toyed with two lines of analysis. The first invokes physics and claims that
on analysis (1) and (2) imply logically incompatible propositions about the
velocity of particles (TLP 6.3751; NB 16.8./11.9.16). A more straightfor­
ward version suggests that on analysis they respectively entail something like
'A reflects mainly light of 620 nm' and 'A reflects mainly light of 520 nm.'
The second line of analysis involves the idea that colours like red are
composed of simpler elements - unanalysable shades of colour. (1) and (2)
are analysed into propositions which ascribe a certain 'quantity' of red and
green respectively to A, plus a supplementary clause stating 'and nothing
else', which means that their conjunction is a contradiction (MS 105; RLF;
PR ch. VHI; BT 473-85).
Unfortunately, as Ramsey detected, both analyses only push the problem
one step back (Mathematics 279-80). The resulting propositions once more
exclude each other; they ascribe one out of a range of incompatible specifi­
cations, a determinate of a determinable. Worse, Wittgenstein realized that
such propositions as (1) and (2) cannot be constructed out of simpler ones,
ascribing degrees of a quality, since logical conjunction cannot reduplicate
the effect of adding degrees. If (1) ascribes to A, for example, 3 degrees of R,
(1*) 4is3R
it cannot be analysed as
(1**) A is 1R.4 is 1 R . 4 is 1R
since (1**) is simply equivalent to 'A is 1R'. Nor can it be analysed as
(1***) A is 1R.4 is 2 R . 4 is 3R
for (1***) contains the analysandum; and '4 is 3R' means either 'precisely
3R', in which case it excludes the other conjuncts, or 'at least 3R', in which
case it entails the other conjuncts. Colour-exclusion thus led Wittgenstein to
realize that statements of degree cannot be analysed to yield ELEMENTARY
PROPOSITIONS which are logically independent.
82
COLOUR
His response was to abandon that requirement, and thereby logical
atomism, which suggests that the Tractatus had conceived of unanalysable
shades of colours as objects, and of elementary propositions along the lines
of (1*). The result is the idea of a 'proposition system': propositions of
degree are compared with reality not individually, but all at once, like the
graduating marks of a ruler; (1) at a stroke determines that A is neither
green nor blue, etc. (WVC 6 3 ^ ; PR 108-11). Secondly, there are non-
truth-functional logical relations: (1) and (2) do not 'contradict', they 'logi­
cally exclude', each other (see TRUTH-TABLES). By this token,
(3) Nothing can be red and green all over
is neither analytic nor empirical (cp. 'Nothing can be white and a metal'), nor
synthetic a priori, as Husserl suggested. We are not prevented from imagin­
ing a counter-example by the transcendental structure of the mind; rather,
nothing would count as red and green all over. (3) is a GRAMMATICAL proposi­
tion, that is, it expresses a rule that excludes as nonsensical a certain combi­
nation of words (WVC 67-8; PR 78-9), namely '4 is red and green all over.'
In his Remarks on Colour (1950-51), Wittgenstein extended this idea. The
INTERNAL relations noticed earlier are part of a whole 'mathematics', 'geo­
metry' or 'logic' of colour (ROC III §§3, 63, 86, 188) which must be dis­
tinguished from its physics. Tractatus 2.0232 was wrong to imply that
determinates of a single determinable share all combinatorial possibilities.
Thus Wittgenstein asks, 'Why is it that something can be transparent green
but not transparent white?' And he insists that such questions cannot be
answered by physical or psychological theories, for they concern not causal
properties of colours, for example that red things emit light of 620 nm, or
irritate people, but their internal properties. He also resists the attempt to
answer them by reference to facts which lie between science and logic, as
with Goethe's 'phenomenological' theory of colours (ROC I §§19, 22, 39-
40, 53, 70-3, II §§3, 16, III §§81-2, 229; WAM 125). He would also reject
the solution offered by scientific realism (see AUTONOMY OF LANGUAGE): to be
both transparent and white is impossible, since to be transparent is to trans­
mit most incident light, while to be white is to reflect most of it, not because
of grammar or empirical fact, but of metaphysical necessity. But no theory
of light transmission or reflection is part of our colour concepts. To reflect
most incident light is not part of the explanation of 'white', and not an
internal property: we would not stop calling clean snow 'white' if measure­
ment revealed that it transmits or absorbs most light.
Wittgenstein himself elucidates the incompatibility between white and
transparent through 'rules of appearance' governing the use of visual terms.
First, something white behind a coloured transparent medium appears in
the colour of the medium (we may know it to be white, but have to use
the medium's colour in representing how it appears); something black
83
 CONSCIOUSNESS
appears black. Consequently, a putative 'white transparent medium' makes
white appear white, black appear black, that is, behaves like a colourless
medium, which is absurd (ROC I §20, HI §173). Second, any coloured
medium darkens the appearance of what is seen through it. For a putative
white medium to do so, it would itself have to be dark, that is, not white
(ROC I §30).
Wittgenstein elucidates other seemingly phenomenological features, for
example that there cannot be a reddish-green, and the contrast between
pure and mixed colours, by reference to standard representations of colour
(colour-octahedron, -circle, -charts), which he characterizes as arrangements
of grammatical rules (LWL 8, 11; PR 51, 75, 276-81). This is linked to the
crucial role of OSTENSIVE DEFINITION: we explain, justify and criticize the use
of our colour vocabulary by reference to samples: 'This colour (pointing at a
chart or a ripe tomato) is red.' Grammatical propositions about colours
reflect normative connections we set up through our employment of colour
samples. For example, we use the ordered pair of a black and a white patch
also as a paradigm of 'lighter' and 'darker', to exclude as nonsensical the
claim that tkis white patch is darker than this black one (RFM 48, 75-6).
The role of ostensive definition explains other features of colour-terms.
(d) They are not denned lexically (a point the Tractatus distorted by hold­
ing that objects cannot be 'described') but through samples (PG 89-
90, 208-9).
(e) Primary colours like red are simple not in the metaphysical sense of
Tractarian OBJECTS, but in that our form of representation teats them
as simple elements of mixed colours, and does not provide a method
of analysing them (RPP I §§605-9).
(f) The blind or colour-blind do not have the colour concepts of the
normally sighted (ROC I §§9, 13, 77, HI passim; RPP I §602; LW II
24-6, 61, 74-9). The reason for this is not that they lack a certain
private experience - the PRIVATE LANGUAGE ARGUMENT undermines the
idea that colours are subjective in that each individual might mean
something different by 'red' in spite of explaining and applying it in
the same way (inverted spectrum). Rather, it is due to the fact that
they lack the perceptual abilities to participate fully in our language-
game with colour-terms. A colour-blind person might know that the
top light of a traffic-light is red, but could not know it by simply look­
ing at the light and saying 'This is red.'
consciousness Wittgenstein's early philosophy involved a form of
SOLIPSISM, according to which reality is identical with life, and life with 'con­
sciousness', that is, my current experience, with the striking result that in
death, when consciousness ends, 'the world does not change but stops exist-
84
CONSCIOUSNESS
ing.' This echoes Schopenhauer, who had claimed that the world is my
representation, and that the concept of representation coincides with that of
consciousness. Although Wittgenstein abandoned his Schopenhauerian meta­
physics after the Tractatus, in his VERIFICATIONIST phase he continued to hold
that 'All that is real is the experience of the present moment' (M 102-3; see
NB 11.6./24.7./2.8.16; TLP 5.621, 6.431; World I §§1, 10, II ch. 1). From
1932 onwards, however, Wittgenstein came to criticize not only this exotic
solipsism of the present moment, but also the INNER/OUTER picture of the
mind as a private realm which has dominated philosophy since Descartes.
'The picture is something like this: Though the ether is filled with vibrations
the world is dark. But one day man opens his seeing eye, and there is light'
(PI II 184). Consciousness is conceived as the ray of light which illuminates
our private mental episodes, an inner glow which, in James's words, marks
'the chasm' separating mind from matter, the foundations of empirical
knowledge from what we can at best infer (LPE 296-7; Psychokgy I 134-6).
In his mature later work, Wittgenstein raised several objections against this
idea of an inner 'world of consciousness' (LW II 21, 74; PI §§412-27; LPE
3 2
°)- 0
(a) The view that the content of consciousness or experience is entities to
which only I have access is challenged by the PRIVATE LANGUAGE ARGUMENT,
which denies that the idea of such private 'thuses' and 'thises' makes sense
(RPP I §§91, 109, 896).
(b) Far from consciousness being known through infallible introspection,
there is no such thing as perceiving or encountering one's own conscious­
ness. If after an accident I tell a doctor 'I am conscious', I do not report the
result of having observed my mind, but simply signal that I have regained
consciousness, something I could equally well have done by saying 'Hello!'
(PI §§416-17; Z §§396, 401-2).
(c) Partly for this reason, it is misguided to seek the essence of conscious­
ness through turning one's attention towards one's own consciousness. What
is needed is an investigation of how the word 'consciousness' and its cog­
nates are used.
(d) Such an investigation reveals that 'consciousness' does not refer to a
phenomenon (state or process) occurring inside us. The alleged ontological
split between the physical world and the world of consciousness is merely a
categorial difference drawn in our language, namely between, those things
which are sentient, that is, capable of perceiving and reacting to their envir­
onment, and those which are not. That healthy HUMAN BEINGS are conscious
(or that they see, feel and hear) is a grammatical proposition, and the sug­
gestion that human beings who behave just as we do might in fact be auto­
matons is absurd (PI §§281-4, 420; Z §395; RPP II §§14, 19, 35; LW II 78).
(e) If this is correct, there is no 'unbridgeable gulf between consciousness
and brain-process', and no unsolvable metaphysical mystery about con-
85
CONSTRUCTIVISM
sciousness (PI §412; BB 47). Although it makes no sense to attribute con­
sciousness to the brain or its parts, and although consciousness is not a pro­
cess which takes place in the brain, there is nothing paradoxical about a
neurophysiological event, whether it be an electrical stimulation of the brain,
or the squeezing of the eyeball, producing certain experiences (e.g., a flash
of light in the visual field). Equally, there is no metaphysical mystery about
the fact that only creatures with a central nervous system of a certain com­
plexity are conscious, although there are scientific puzzles here which Witt­
genstein did not address, for example, why and in what way capacities for
sensation and volition presuppose certain neurophysiological mechanisms
and processes, and how these capacities emerged in the evolutionary pro­
cess.
constructivism see GENERAL PROPOSITIONAL FORM; MATHEMATICAL PROOF;
NUMBERS
contextualism This doctrine holds that in the explanation of meaning,
judgements, sentences or propositions take priority over concepts or words.
It is implicit in Kant (B92-3), who held that the sole function of concepts is
to be employed in judgements, and Bentham, who gave a contextual defini­
tion of grammatical particles like 'if or 'but', that is, explained them by
paraphrasing the sentences in which they occur. Frege insists on the pri­
macy of judgements (thoughts) over concepts: rather than constructing jud­
gements out of concepts (like traditional logic), he derives the latter from
analysing the former. This idea underlies his famous 'context principle':
never 'enquire after the meaning of a word in isolation' and 'Only in the
context of a proposition do words mean something' (Foundations x, §§60—2,
106; see Posthumous 15-16, 253). This principle has three implications.
Firsdy, a sufficiency condition: for a word to have meaning it suffices that it
play a part in expressing a judgement. This allowed him to insist, against
psychologism, that for a sentence to be meaningful, it is not necessary for
every individual word to be associated with an idea. Secondly, composition-
alism: the meaning of a word is its contribution to the content of sentences
in which it occurs, because the latter is composed of the meanings of its
constituents (similarly, for Russell a proposition is the value of a proposi­
tional function). Finally, a restrictive condition: only in a sentence which
expresses a judgement have words a real logical 'content', for only there do
they (partly) determine the validity of inferences.
When Frege divided content into 'sense' and 'meaning', he regarded the
sense and the meaning of a sentence as determined respectively by the
senses and meanings of its constituents. But he rejected the legitimacy of
contextual definitions, and never adapted the context principle to the dis­
tinction (Posthumous 255-6; 'Sense' 35-6; Laws LI §66). The Tractatus does
86
CONTEXTUALISM
this, but in a modified form. Whereas Frege distinguished between the
'saturated' names of objects (e.g., 'Paris') and the 'unsaturated' names of
functions (e.g., 'is the capital of France'), Wittgenstein insists that all names
are unsaturated, that is, have meaning only in coordination with others.
Moreover, he denies that propositions have MEANING, and that NAMES have
sense. 'Only the proposition has sense. Only in the context of a proposition
has a name meaning' (TLP 3.3, see 3.314). Two rationales for this 'restric­
tive principle' concerning names can be detected. The explicit one (TLP
2.0121-2.0131; PT 2.0122) derives from the isomorphism between language
and reality postulated by the PICTURE THEORY. AS regards their combinatorial
possibilities, names behave like the objects they stand for. It is essential to
OBJECTS that they are concatenated with other objects in facts: an object
cannot occur on its own, but stands in determinate relationships with other
objects (see LOGICAL SPACE). In going proxy for objects, names do the same,
hence they are part of FACTS of a special kind, namely propositions. What
depicts the fact that a stands in the relation R to b is not a mere list of signs,
but the fact that 'a' stands to the left and 'b' to the right of 'R' (TLP
3.143 If.; see ELEMENTARY PROPOSITION).
The implicit rationale is provided by the compositionalism of the Tracta­
tus. In line with a (possibly independent) suggestion by Frege, it seeks to
explain the 'creativity of language', the fact (first noted by von Humboldt)
that we can understand propositions which we have never heard before
(TLP 3.318, 4.02-4.03; NL 98, 104; Posthumous 225, 243; Correspondence 79).
The Tractatus's solution has been widely accepted by contemporary philoso­
phers of language. All that is needed to understand an unlimited number
of propositions is knowledge of the primitive expressions (names), and their
combinatorial rules. To understand a proposition is to understand the
meanings of its constituents plus its LOGICAL FORM, that is, the manner of its
composition (contemporary theories add that the rules of composition are
recursive, and hence allow the formation of an infinite number of sen­
tences). At the ultimate level, the sense of an elementary proposition is a
function of its constituent names, that is, both of their meanings — the
objects they represent - and of their logical form, their combinatorial possi­
bilities. By the same token, the role of names is to contribute to the deter­
mination of the sense of elementary propositions. Alas, both rationales show
at best that names must be capable of occurring in propositions, not that
they have meaning only when they actually occur in propositions, as the
restrictive principle requires.
Wittgenstein later rejected both the idea that PROPOSITIONS must consist of
function and argument, and compositionalism generally. The sense of a pro­
position is not determined exclusively by the meanings of its constituents
and the mode of their combination, but depends at least partly on its role,
on how it is used on a particular occasion of utterance. UNDERSTANDING the
87
 CONTEXTUALISM
components and mode of composition of a sentence may be a necessary
condition for understanding it, but it is not sufficient. Given our method of
determining time by reference to the sun's zenith, 'It is 5 o'clock on the sun'
of itself makes no sense (PI §§350-1; BB 105-6; RPP II §§93-4). We could
stipulate a sense, but that is not the same as calculating it from the senten­
ce's components and mode of composition. Moreover, logical form and
status are given by grammatical form only in so far as a given type-expres­
sion form is canonical^ used for a certain purpose. And the standard purpose
of a type-sentence may be at odds with what its linguistic form suggests. It is
common practice to give orders with declarative or interrogative sentences
('I'd like you to shut the door', 'Would you shut the door?'), or to ask ques­
tions with imperatives ('Tell me what you think!'). Whether a sentence is a
GRAMMATICAL proposition, that is, typically expresses a linguistic rule, depends
on its role or function within our linguistic practice: 'War is war' is typically
not used to express the law of identity. Finally, on a given occasion a form
of words may serve a non-canonical purpose — as with a rhetorical question.
This depends on how the speaker uses it on that occasion, and is evident
from how he would explain or defend his utterance, and what responses he
would admit as relevant (PI II 221; LW I §17; MSI31 141-2; see INTENDING
AND MEANING SOMETHING).
This functionalist account of sense removes an objection to the idea that
first-person psychological statements like 'I am in pain' are AVOWALS rather
than descriptions or reports, namely that they must have the same (descrip­
tive) sense as third-person statements ('H.G. is in pain'), since they combine
equivalent components in the same way (LW I §44). It also means that whe­
ther a combination of signs is NONSENSE is no longer decided by reference to
general rules alone, but also depends on the circumstances in which the
expression functions (PI §489; OC §§229, 348-50, 433); and it supports
Wittgenstein's warning that focusing on the form of expressions rather than
their use leads to philosophical confusion (LC 2; AWL 46; PI §§10-14).
Finally, it challenges an assumption of truth-conditional semantics, namely
that sentences have literal meanings determined only by their components
and logical form.
Wittgenstein continued to condone the general idea of the primacy of
propositions over their constituents. However, the rationale is novel. What
gives meaning to words is no longer their being embedded in a logical form,
but their being incorporated in a language-game (see USE). 'A word has
meaning only as part of a sentence' (PI §49). Taken at face value this is
wrong. When I address someone, or paint ' W C on a door, I have not
made a meaningless utterance, or written down a meaningless mark. But
with occasional exceptions (PR 59), Wittgenstein explains his dictum in a
way which acknowledges that individual words can mean something without
actually occurring in a proposition. What he insists on is that a word must
88
CONTEXTUAUSM
be capable of occurring in propositions, and that such occurrences of words
are semantically primary. He reaches this conclusion through the following
(implicit) steps.
(a) A proposition is the minimal unit by which a move is made in the
language-game: only propositions can say something. There are no
half-propositions in the sense in which there is half a loaf of bread
(BT 1; PG 44).
(b) Naming an object is no more a move in a language-game than put­
ting a piece on the board is a move in chess. Naming presupposes a
sentential context in that it is essentially a 'preparation' for sentential
use (PI §49; PLP 13-14, 199, 318-20).
(c) A name can be used in isolation only if there is a LANGUAGE-GAME in
which such moves are made: ' W C would riot be a label if we did not
talk about lavatories.
(d) Understanding a word implies inter alia knowing how to use it in sen­
tences.
The kernel of truth in Frege and the Tractatus is*that the meaning of a word
is determined by how it can be used within sentences. But it does not follow
that the word has a meaning only in the context of a sentence. On the con­
trary, it is the individual word which has such a use and hence a meaning. If
it is clear what role a word would play in a proposition, it has a meaning
whether or not it actually fulfils that role.
This position naturally leads to an extension of contextualism to the idea
that sentences have meaning only within the context of a whole language
(see FORM OF LIFE). A proposition can be a move only in the context of the
whole 'game of language'. 'To understand a sentence means to understand
a language' (PI §199; see PG 172; BB 5; LW I §913). This semantic holism
is reminiscent of Quine and Davidson. Taken literally, it implies that one
cannot understand any part of a language unless one understands every
part, which ignores that we have to learn a language by segments, and that
there are degrees of understanding. On a charitable interpretation, it means
that one cannot credit someone with understanding just one sentence and
nothing else. For a proposition is a sign in a system, one possible combina­
tion of words among, and in contrast with, others. Hence, understanding a
proposition is part of the 'mastery of a technique' (PI §199; PG 63, 152-3).
It involves both the ability to employ a word in some other contexts, and
knowledge of some of the logical links between the given sentence and
others. 'Light dawns gradually over the whole' (OC §§141-2). We do not
learn everything at once, but our grasp of each part is complete only once
we have mastered the whole. Thus understood, semantic holism explains
rather than ignores the fact that there are degrees of understanding.
89
 CONTRADICTION
contradiction For Wittgenstein, a contradiction like 'p.~p' is on a par
with a TAUTOLOGY like '~{p.~p)' in that it is not nonsensical but senseless,
because it says nothing. By contrast, the law of contradiction is not the vac­
uous '~{p~p)', but a rule which prohibits an expression like '/>.~/>'. What
logicians are afraid of are not contradictions per se, which have a legitimate
role, notably in reductio ad absurdum arguments, but violations of this rule, for
example a failure to withdraw a postulate which implies a contradiction.
There is no such thing as a contradictory rule, since it could not tell one
what to do, and a contradictory proposition is no more a move in the
language-game than placing and withdrawing a piece from a square is a
move in chess (WVC 130-1, 176, 199-200; PG 128-9, 305; AWL 4; LFM
209, 212-14, 223; RPP I §44; RPP II §290).
Wittgenstein's remarks on the consequences of contradictions for the
foundations of mathematics are self-consciously provocative. He does not
tolerate, let alone promote, contradictions. However, he regarded as super­
stitious the sceptical fear that there might be 'hidden' contradictions which,
like a germ, might infect the whole body of mathematics unbeknownst to us,
and consequendy he rejected the idea which underlies Hilbert's META-
MATHEMATICS that such contradictions should or could be precluded in
advance by consistency proofs (WVC 119; RFM 204-19, 254-6, 370-8,
400-1, 410; LFM 7-8, 67, 209-30). A hidden contradiction is not an unno­
ticed contradiction, that is, one which is explicit in a set of rules, and has
merely been overlooked, or one which can be generated according to an
established method (WVC 120, 143, 174-5, 208; LFM 226). Rather, it is
one which. is added to the system by a new, unforeseen type of construction
- such as the construction of statements like 'X is a member of itself.'
According to this distinction, Russell did not discover an existing contra­
diction in Frege's calculus, but invented a way of constructing a contra­
diction, and thereby modified that calculus. Nothing forces us to accept this
kind of modification. We can decide that the path leading to the contra­
diction is not a proper derivation within the system. The rules we have
operated commit us only to what can be generated through their straightfor­
ward application, not to what can be added. Equally, no meta-mathematical
discovery could produce a system immune to the possibility of such con­
structions. Certainty of this kind could only be achieved by a 'good angel'
(RFM 378; LFM 221-4; see MATHEMATICAL PROOF).
When a contradiction is constructed or detected, it does not show that
everything we did before was wrong. A contradiction is harmful only if it
brings the application of the calculus to a halt. Thus, if it were suddenly
noticed, because, say, the vice-president turns up for the first time, that cer­
tain statutes made inconsistent demands on the seating of the vice-president
at state banquets, this would not show that what we did previously was
wrong (LFM 210). By the same token, it is difficult to see how our basic
90
CONTRADICTION
arithmetic could be overturned by recherche discoveries in mathematical
logic, a difficulty which lends support to Wittgenstein's suggestion 'that Fre­
ge's and Russell's logic is not the foundation of arithmetic anyway' (LFM
228; WVC 149; RFM 400-1). However, Wittgenstein himself insisted that
an expression like ' $ and do not <I> in situation X' is not what we call a rule
(PG 305). Hence, one should add that in the case of our unnoticed contra­
diction there was something wrong before it was revealed in our practice,
not with what we did, but with the statutes, namely that they did not pro­
vide coherent guidance on the seating of the vice-president. Equally, an
arithmetic which did not prohibit division by 0 would be inadequate even
before someone started dividing by 0.
Even if one accepts Wittgenstein's ideas about unnoticed and hidden con­
tradictions, some of his remarks on what to do once a contradiction comes
to light are problematic.
One might say 'Finding a contradiction in a system, like finding a germ
in an otherwise healthy body, shows that the whole system or body is dis­
eased.' - Not at all. The contradiction does not even falsify anything. Let
it lie. Do not go there. (LFM 138) 0
This is like saying that we can avoid trouble with the statutes about seating
arrangements simply by refraining from having state banquets; it defies the
purpose of the rule system. In other passages, Wittgenstein takes a more
plausible fine: when we discover a contradiction, remedial action is called
for, but it can always be provided, notably through ad hoc stipulations like
Russell's, which prohibit expressions like 'X is a member of itself' Thus, the
main problem with contradictions is that anything would follow from a con­
tradiction, but we can avoid this by making it a rule that no conclusions
should be drawn from a contradiction (WVC 132; RFM 208, 373-7; LFM
209-10, 219-28).
Waismann and Turing protested that this only cures the symptoms, since
an inconsistent system will create indefinitely many contradictions. In reply
it has been argued that we can only derive indefinitely many contradictions
by (tacitiy) drawing inferences from a contradiction, which would mean that
Wittgenstein's rule would prevent the derivation of contradictions. But Witt­
genstein himself acknowledged that the contradiction can be contained only
if we can survey the system, which means that ultimately the solution to the
emergence of a contradiction is to disentangle ourselves from the confusions
engendered by our own rules. Once we have done so, the straightforward
solution is to modify the system, for example, by declaring one of two con­
flicting rules obsolete (RFM 209; PI §125; LFM 210).
Wittgenstein also rejected Turing's suggestion that bridges might collapse
as the result of a hidden contradiction in our mathematics (LFM 210-21). If
a bridge collapses, either our physics is wrong, or we have made a mistake
91
 CONVENTIONALISM
in the calculations. Strictly speaking this is wrong, because in an inconsistent
system someone might argue 'p .~p; ergo 2x2- 369' and use that result in
constructing a bridge. But Wittgenstein is right that we would not call this
calculating, and that the real problem here is not the contradiction, but the
drawing of such absurd conclusions.
Wittgenstein's general attitude towards the fear of hidden contradictions
also inspires his ill-famed discussion of Godel's first incompleteness theorem
(RFM 116-23, 383-9). The theorem states that for any axiomatic system S
adequate to formalize arithmetic - such as that of Principia - there is at least
one well-formed formula which cannot be proven in this system. The tech­
nique used in reaching this result is to translate meta-mathematical state­
ments about the provability in S into arithmetical statements, which are
themselves part of S. On that basis we can prove within S an arithmetical
statement 'P' which represents within S the meta-mathematical statement 'P
is not provable in S\ or, more vividly, 'I am unprovable.' But if S is con­
sistent, P is true (no false proposition can be proven), and hence unprovable.
Wittgenstein did not impugn the validity of the proof, but only the inter­
pretation of T ' as saying of itself both that it is unprovable and that it is
true. One of his arguments is that this interpretation is paradoxical, since for
'P" to be true in S is for it either to be an axiom of S or to have been proven
from those axioms. Critics have also detected another line of argument,
namely that Godel's interpretation of 'P' is untenable because it is on a par
with the liar's paradox. This would ignore that in Godel's proof no single
self-referential statement like 'I am false' occurs; rather, we have two ver­
sions of the same proposition in two different systems, one version is true
but unprovable in S, the other is true and provable in the meta-mathema­
tical system M. However, Wittgenstein's real point is precisely that there
cannot be two versions of the same mathematical proposition in two differ­
ent systems, since a mathematical proposition has sense only as part of a
particular proof-system. According to him, Godel's proof has in fact con­
structed two different propositions, one of which - '/" in S - is unprovable,
l
while the other - P is unprovable in S' - is true but part of M, and hence
has no sceptical implications. Neither line of argument evidences the egre­
gious technical incompetence of which Wittgenstein has been accused. But
both presuppose his view that a sentence is a meaningful mathematical truth
only if it has been derived within a particular system of MATHEMATICAL PROOF.
Without independent reasons for this view, Wittgenstein's attack on Godel is
question-begging, since Godel's interpretation implies precisely that there is
a gap between mathematical meaning and truth on the one hand, and
mathematical proof and provability on the other.
conventionalism see FORM OF REPRESENTATION; LOGIC; MATHEMATICAL PROOF
92
CRITERIA
criteria These are ways of telling whether something satisfies a concept X,
or evidence for something's being an X. Although used by Plato, the term
has achieved philosophical prominence only through Wittgenstein's later
work. It has been treated as a technical term, in spite of its modest fre­
quency and the fact that there is only one passage which defines it. For the
most part, Wittgenstein's use of the term is in line with the ordinary one.
But sometimes he specifies that criteria provide a special kind of evidence.
The root of this idea lies in a distinction of his VERLFICATIONIST phase (WVC
97-101, 159-61, 210-11, 255-6; PR 94-5, 282-6; LWL 16, 66; M 55-61;
PG 219-23). A 'genuine' PROPOSITION must be conclusively verifiable by
reference to reality, which leaves as candidates for genuine propositionhpod
only sense-datum propositions that describe immediate experience. State­
ments about material objects or the experience of others cannot be so ver­
ified, and are mere 'hypotheses'. They are not strictiy true or false, and not
propositions in the same sense as 'genuine' ones, but rather rules for the
construction of such propositions. Those propositions which give evidential
support to hypotheses are called 'symptoms'. Thus, the different views of a
material object support hypotheses about the material object itself; the result­
ing hypotheses explain our previous experiences ancf predict future ones ('If
viewed from a different angle, the object will look like this'). The relation
between hypothesis and evidential symptom falls short of entailment: symp­
toms never conclusively verify or falsify a hypothesis, they only make it more
or less probable. For the evidential support is defeasible: the addition of
further propositions to the set of symptoms may make the hypothesis less
plausible. Moreover, a hypothesis can always be upheld or abandoned by
adopting auxiliary hypotheses. Which course we take depends on considera­
tions of simplicity and predictive power. That relation is nevertheless 'gram­
matical' or 'logical': what symptoms render more or less probable what
hypothesis is determined a priori, not through experience (induction).
Wittgenstein later realized that while his candidate 'genuine propositions'
do not in fact allow of verification, since they are not descriptions but
AVOWALS, his so-called 'hypotheses', humdrum propositions like 'The table is
round' or 'She has a toothache', are not rules or laws, and can sometimes be
conclusively verified. As a result, the relation between a hypothesis and its
evidential symptoms was replaced by that between a proposition and its
'criteria' (first in his Cambridge lectures of 1932-3: AWL 17-19, 28-35, 59-
62). Like symptoms, criteria are grounds or reasons fixed by grammar, not
experience. But there are also differences between the earlier and the later
relation, (a) The relata of the criterial relation are characterized diversely:
'phenomena', 'facts' and 'propositions' are criteria of 'sentences', 'state­
ments', 'phenomena', 'facts', 'knowledge', 'assertions', 'concepts' and 'words'.
Ultimately, these variations are terminological; they express a single idea in
a formal (linguistic) or material mode, and by reference to propositions or
93
CRITERIA
concepts. The basic point is that certain phenomena or facts license the
application of certain words, (b) Criteria can conclusively verify a proposi­
tion (see below), (c) They may be unique, although many concepts have
multiple criteria.
In the Blue Book, Wittgenstein gives an explicit explanation which reverses
his initial terminology (BB 24—5). 'Symptoms' are now denned as empirical
evidence; they support a conclusion through theory and induction. By con­
trast, a 'criterion' q for a claim p is a ground or reason for the truth of p, in
virtue not of empirical evidence, but of grammatical rules. It is part of the
meaning of p and q that q's being the case - the satisfaction of the criterion
- is a ground or reason for the truth of p. An inflamed throat is a symptom
of angina; the presence of a particular bacillus is the 'defining criterion'.
Commentators have sometimes given the impression that for Wittgenstein
the use of all propositions and concepts is governed by criteria. In fact, he
stated that avowals and concepts which are denned through OSTENSIVE
DEFINITION (e.g., colour predicates) are not subject to criteria. The same
might hold for FAMILY-RESEMBLANCE concepts. On the other hand,
expressive behaviour is a criterion for third-person psychological utter­
ances;
performances are criteria for potentialities, powers and abilities (notably
applying and explaining a word correcdy are criteria for UNDERSTANDING
«); _
scientific concepts like angina are governed by criteria, although these
often fluctuate (see below);
mathematical concepts are governed by 'defining' criteria (having three
sides is the criterion for a plane figure's being a triangle);
MATHEMATICAL PROOFS are criteria for mathematical truths, and the result
of an arithmetical operation is a criterion for its having been carried
out (RFM 319) (if you do not get 144, you have not squared 12); and
applying count-nouns requires 'criteria of identity' (see PRIVATE LANGUAGE
ARGUMENT).
The idea of criteria has three distinctive and problematic features:
(a) Criteria determine the meaning of the words they govern. What connects
meaning and criteria is verification (AWL 17-19, 27-8; PI §353): to explain
one's criteria for something's being F is to specify how 'a is F is verified.
The meaning of F is not necessarily given by specifying the criteria: 'being in
pain' does not mean 'screaming in circumstances of injury'. Nonetheless, the
criteria determine (at least partly) the meaning of F. To specify the criteria for
F is to specify rules for the use of F, and hence (partially) to explain its
meaning. Criteria are 'fixed by grammar', 'laid down by language' and in
94
CRITERIA
that sense a matter of 'convention' (AWL 28-9; BB 24, 57; LPE 293; PI
§§322,371).
Accordingly, the relation between concepts and their criteria is INTERNAL.
It makes no sense to say, for example, 'Here is pain, and there is behaviour
- it just happens that they are associated' (LPE 286; LSD 10). It also means
that a change in criteria is a conceptual change, a change in the meaning of
words: that q is a criterion for being F is partly constitutive of the concept of
being F. Thus, mathematical proofs are concept-forming, since they lay
down new criteria for applying, for example, • numerals. And scientists often
change the meaning of words under the impact of empirical findings, for
example when they discover that one phenomenon from a certain cluster
causes the others (as in the case of angina where the bacillus causes the
symptoms) or allows of precise measurement (Z §438).
Criteria have since been invoked, firsdy to combat scepticism about other
minds, and secondly to develop an anti-realist theory of meaning, which, in
contrast to the alleged realism of the Tractatus, is based on assertability-
conditions rather than truth-conditions. The latter application is unfaithful
to both the early (see VERIFICATIONISM) and the later Wittgenstein, whose con­
ception of PHILOSOPHY rules out theories of mearSng; and, as we have seen,
only some uses of language are subject to criteria.
Distortions apart, Wittgenstein*s treatment of criteria has come in for
powerful criticism. Radical empiricists like Quine deny that there is such a
thing as conceptual evidence or internal relations. Others have claimed that
criteria cannot be a matter of convention: no one has ever stipulated that
behaviour should express pain, and no one could decide otherwise. Propo­
nents of a realist semantics Like Putnam add that the criteria we use to
establish whether something is, for example, a case of angina are only crude
ways of detecting a natural kind. What 'angina' means is determined by the
ultimate scientific theory about the matter (see AUTONOMY OF LANGUAGE). By
that token, the idea that cases in which scientists adopt new criteria for the
application of a term like 'angina' amount to cases of conceptual change is
wrong because it implies that we are no longer talking about the same
thing. Arguably, however, Putnam's objections illustrate an important lesson
of Wittgenstein's account, namely that there is a 'fluctuation in grammar
between criteria and symptoms' (PI §354, see §79). The logical status of cer­
tain relations may change from being criteria! to being symptomatic, and it
may do so as a result of empirical discoveries (Wittgenstein may have
derived this lesson from his work on shock during the war). Scientific con­
cepts are typically held in place by several criteria, and we can abandon
some of them while retaining others. That is why we are not simply talking
about a completely different phenomenon. Nevertheless, to alter the criteria
is to alter how we explain and employ, for example, 'angina', and hence
amounts to a modification of our concept.
95
 CRITERIA
(b) Criteria are ways of telling how one knows something. To specify the cri­
teria for the truth of p is to characterize ways of verifying p, of answering
the question 'How do you know?' (AWL 17-19, 28; BB 24-5, 51, 57; Z
§439; LPE 293; PI §§182, 228). In line with his earlier conception of symp­
toms, Wittgenstein sometimes refers to criteria as 'evidence'. This is mislead­
ing, because it suggests that p is logically independent of q, whereas the
relationship is in fact an internal one, yet it does righdy suggest the defeasi-
bility of (some) criteria (see (c) below). But it is important to stress that cri­
teria differ from necessary and sufficient conditions not just by virtue of
being (in some cases) defeasible but also in that they must be features which
can be invoked to justify the application of a term. There may be necessary
and sufficient conditions which do not satisfy this condition. Thus, Wittgen­
stein claimed that being bivalent is necessary and sufficient for being a pro­
position, but not an independent feature by which we could recognize
something as a proposition (see PI §136; BIPOLARITY).
(c) The criteria for some words are defeasible. This legal term is not used by
Wittgenstein, but it indicates the special nature of criterial evidence. In some
cases, a criterion is a logically sufficient condition, or even a necessary and
sufficient condition, for something's being X: the presence of a certain bacil­
lus for angina, having three sides for being a triangle. In other cases the cri­
terial relation does not amount to entailment, but shares a feature of
inductive evidence: it need not be decisive, because it can be undenriined
by further evidence. Whether or not criteria conclusively support p may
depend on the circumstances. If Susan screams 'It hurts' and writhes on the
floor, this is a criterion for her being in pain; but if she is rehearsing for a
play, it will not confirm that she is suffering. Such defeasibility cannot be
avoided by claiming that a criterion q is a necessary constituent of a suffi­
cient condition which includes those circumstances which, together with q,
entail p. For there is no definite list of such circumstances, and even if there
were, it is not part of our explanations of psychological terms, and hence
not part of their meaning (Z §§117-22).
Defeasibility threatens to open the flood-gates to scepticism about other
minds. For all our criterial evidence, we may be wrong in our inferences
from it about Susan's state of mind. It has therefore been suggested that cri­
teria for psychological terms should not be considered as evidential (in an
inferential sense). If we see her screaming and writhing, we do not infer
(consciously or unconsciously) that she is in pain from behavioural evidence,
we simply register her agony. This interpretation is in line with Wittgen­
stein's attempt to avoid the INNER/OUTER picture of the mind as something
hidden for which we have merely evidence. The answer to the question
'How do you know she was in pain?' is simply 'I saw her writhing in agony.'
Like straightforward observations of material objects, this does not adduce
96
CRITERIA
evidence, but specifies a perceptual capacity which direcdy shows us how
things are. This perceptual model makes our relation to other minds less
open to sceptical challenge, while still allowing for the possibility of error. It
also acknowledges that what we observe is not colourless movements descri­
bed in neutral, physical terms, but /ratn-BEHAViouR. On the other hand, it
does not work for cases like 'Helga intends to go to London.' Here, the
answer to the question 'How do you know?' is not simply 'I saw her', but,
for example, 'She told me, and later bought a ticket.'
In any event, Wittgenstein accuses the sceptical challenge of ignoring the
internal relation between psychological concepts and behavioural criteria,
and thereby distorting the concepts involved. The fact that criterial evidence
is defeasible does not entail that it is actually defeated in a particular case.
Any challenge in a particular case must point out defeating conditions, but
these are themselves defeasible, and quickly peter out (RPP I §137): there is
no more room for doubt once it turns out that Susan has broken her leg
during the rehearsal (see SCEPTICISM). The possibility of lying and pretence
does not overturn this verdict. For one thing, the very concept of pretending
to be in pain is parasitic upon the concept of being in pain; it makes sense
only because there are manifestations of pain wffich are not subject to pre­
tence, such as the grimace of an infant. Moreover, there are criteria for pre­
tending no less than for being in*pain. One can't pretend to be distraught
while throwing oneself off the roof. Doubt in such circumstances betokens
not caution, but misunderstanding or distortion of the concepts involved. In
such circumstances, we can know, and be certain, and a 'proof or 'guaran­
tee' is provided by behavioural criteria (PI §§246, 249-50, II 181, 222-9;
LPE 293; Z §§570-1).
Wittgenstein's last writings question the idea of criterial support as deci­
sive grounds. There can be no proof of third-person ascriptions of emotions,
and we may often be unable to decide whether someone is, for example,
annoyed. But this does not hold for sensations, and does not rehabilitate
scepticism. For this 'indeterminacy' and 'unpredictability' is constitutive of
some of our concepts of the inner. Moreover, often those who are closely
acquainted with a person can make even the most subde emotional ascrip­
tions with certainty, without being able to specify conclusive criteria, since
their evidence is 'imponderable', that is, consists of a syndrome of beha­
viour, context and previous events (PI II 227-8; LW II 70, 87, 90-5). Cri­
teria are neither the linchpin of a new semantics nor the wonder weapon
against scepticism about other minds. But they signify conceptual connec­
tions between psychological concepts and behaviour which are unwisely
ignored by adherents to the inner/outer picture of the mind.
97

D elements. But such a configuration is something absolutely precise: either it
determinacy of sense Frege had postulated that a concept must have
'sharp boundaries', that is, that its definition must 'unambiguously deter­
mine, as regards any object, whether or not it falls under the concept' (Laws
II §§56-64; Posthumous 155). A concept without a precise definition is not a
genuine concept. One rationale behind this is the principle of bivalence:
every sentence must be determinately true or false. Another is that Frege
treats concepts as functions, and a mathematical function is defined only if
its value is stipulated unequivocally for every argument. Finally, for Frege
the sense of a complex expression is a function of the senses of its con­
stituents, which means that indeterminacy is contagious. To avoid vague­
ness, , a definition must be complete: it must determine for any object
whether or not it falls under the concept, whatever the facts.
Wittgenstein imbibed Frege's ideal of determinacy of sense and the
demand for completeness of definition. But whereas for Frege and Russell
the vagueness of natural language is a defect which must be avoided by an
ideal language suitable for scientific purposes, for the Tractatus it is a surface
phenomenon, that is, a phenomenon which analysis reveals to be merely
superficial. Many sentences of ordinary language appear vague or ambig­
uous. However, this vagueness 'can be justified' - ordinary language is in
good logical order. Although a proposition may leave certain things open, if
so it must be determinately ^determinate, that is, it must be settled what
precise range it leaves to the facts. 'The watch is lying on the table' leaves
open the precise location of the watch. But it must define absolutely sharply
what possible locations it can occupy. Hence, LOGICAL ANALYSIS reveals it to
be a statement to the effect that there are two objects of such-and-such a
kind which stand in one out of a variety of possible spatial relations to each
other. Even this may create problems, since it might be unclear what pre­
cisely counts as lying on the table. Nevertheless, Wittgenstein insists that
what I mean by uttering that sentence on a specific occasion must always be
perfecdy sharp. The implications of a given proposition 'must be settled in
advance' by its sense (TLP 3.24, 5.156; NB 7.9.14, 16.-22.6.15; PT
3.20101-3.20103). Determinacy of sense is a precondition of there being
any sense at all.
98
DETERMINACY OF SENSE
Wittgenstein shares Frege's commitment to bivalence: 'A proposition
must restrict reality to two alternatives: yes or no' (TLP 4.023; FW 55). But
that commitment is in turn derived from the PICTURE THEORY: the sense of a
proposition is a state of affairs, that is, a possible configuration of simple
exists, or it does not. A proposition must be determinate because there
must be a precise configuration of simple elements which either verifies or
falsifies it. The logical requirement that the sense of propositions be deter­
minate mirrors the metaphysical nature of facts, and implies that the analy­
sis of all propositions terminates with logically proper NAMES which stand for
indestructible simple OBJECTS. A proposition can depict a precise configura­
tion of elements only if its ultimate constituents stand in a one-to-one cor­
relation with these elements. Otherwise, the FACT that its ultimate
constituents are combined in a certain way does not depict a specific com­
bination of things.
After the Tractatus, Wittgenstein's attitude to mdeterminacy changed. He
abided by the conviction that 'all the propositions of our everyday language,
just as they stand, are in good logical order' (TLP 5.5563). 'Ordinary lan­
guage is all right' (BB 28; PI §98). But his conception of that order changed
radically, as he abandoned the idea that speaking a language is operating a
CALCULUS according to definite rules. Not only is it incoherent to suppose
that every aspect of language must be governed by rules, it is equally mis­
guided to insist that the rules that do operate must preclude the possibility
of vagueness under all conceivable circumstances. At first, he held fast to the
idea that the logical order of language mirrors the structure of reality, while
modifying his atomistic ontology. Inexactness or vagueness, he argued, is an
mtrinsic property of certain objects and experiences. It distinguishes, for
example, the geometry of the visual field from Euclidean geometry, and is
essential to memory images and some visual experiences. The ,'inexact'
terms of ordinary language are best suited to express exactiy the 'blurred-
ness' of what we experience (WVC 55-6; PR 260-3; PLP 208-11).
Philosophical Investigations is often seen in the same light, namely as main­
taining that vagueness is an essential feature of language. Thus understood,
it has been a major inspiration behind attempts to construct a logic of
vagueness. However, Wittgenstein's mature treatment of the topic (PI §§75-
88, 98-107) does not promote vagueness; it merely resists the dogmatic
demand of determinacy of sense, that is, the insistence that the possibility of
doubt or disagreement about the application of an expression must be elimi­
nated. Equally, Waismann's influential term 'open texture' rejects not exact­
ness, but the demand that inexactness should be impossible (although on the
basis of VERIFICATIONIST ideas Wittgenstein had repudiated by the time of the
Investigations). Not all concepts are actually vague, and though most empirical
concepts allow of borderline cases, they do not thereby become useless, an
99
DETERMINACY OF SENSE DETERMINACY OF SENSE

idea Hart extended to legal concepts in order to combat legal formalism nate: there may be borderline cases, but it must be exactiy determined what
and rule-scepticism. counts as such a case. However, this idea leads to a vicious regress. If we try
Wittgenstein rejects the assumptions behind the demand for deteixninacy. to make the limits of an area more precise by drawing a line, that line has a
Bivalence and BIPOLARITY are optional features of language. Moreover, breadth. If we try to avoid this by using the colour-edge of the line, the only
vagueness is not necessarily contagious, as the compositionalism of Frege way of determining what counts as overlapping this exact boundary is to
and the Tractatus insisted. A claim that the river bank is overgrown by plants draw another line, etc. (PI §88; Z §§441-2).
is not mdefinite because biologists may puzzle over whether to classify cer­
tain micro-organisms as plants or animals. A resolution of such puzzles
through a sharp definition of 'plant' would not sharpen our understanding
of every sentence in which the word occurs; it would introduce a new con­
cept (BT 69, 250; MSI 15 41).
Far from insisting on the desirability of vagueness, Wittgenstein insists that
'inexact' and 'incomplete' are terms of reproach, 'exact' and 'complete'
terms of praise. But he takes Frege and the Tractatus to task for distorting
the ideal of exactness.
(a) There is no single ideal of exactness. The contrast between exact and
inexact is relative to a context and a purpose (e.g., whether we are measur­
ing our distance to the sun or the length of a table) (PI §§88, 100; BT 249-
50). An inexact definition is not one which fails to meet the elusive ideal of
determinacy, but one which fails to meet the requirements of understanding
in a given context.
(b) No EXPLANATION could avert all possibility of mdeterminacy, since no
system of rules can budget for the coundess bizarre possibilities in advance
(PI §§80, 84-7).
(c) Although vagueness is a defect, a proposition with a vague sense still
has a sense, just as a vague boundary still is a boundary. If there is only one
gap in an enclosure, it is determined that there is only one way out (a fly-
bottle is a trap, although there is a way out). If I tell someone 'Stand
roughly there', accompanied by pointing at a particular spot, some actions
will count as complying with the order, and others not, although there may
be borderline cases. For a concept to be useful, all that is required is that it
is defined for some cases, so that some things would definitely fall under it,
and others definitely would not. The Sorites paradox arises out of the failure
to recognize that this absence of precise bounds is constitutive of perfecdy
useful concepts like 'heap': the order to make a heap is clear, although the
order to make the smallest heap which still counts as such is not (PI §§68-
71, 79, 88, 99; PR 264; PG 236-40). 'Heap' is not the kind of concept to
which one can apply mathematical induction. By the same token, names
need not be analysable into a set of uniquely identifying descriptions to have
a use, and a FAMILY-RESEMBLANCE concept like 'game' does not cease to be a
concept by failing to be defined analytically.
(d) One might respond in the spirit of the Tractatus that although the rules
may allow a certain degree of elasticity, that degree must itself be determi-

100 101

E (b)
elementary proposition (Ekmentarsatz) Traditional grammar regarded
subject-predicate sentences like 'Mary is blond' as simple. Logical atomism,
by contrast, used modern logic to show that such propositions are 'mole­
cular', that is, truth-functions of simpler propositions, just as 'p.q' and
'p'Dq' are truth-functions of'/?' and 'q'. 'Atomic' or 'elementary' proposi­
tions are 'the simplest' propositions into which all others can be analysed
but which cannot themselves be analysed into simpler propositions (RUL
8.12; NL 95-7; NM 111). For Russell, the foundations of knowledge also
provide the foundations of linguistic meaning. According to his empiricist
'principle of acquaintance', every proposition which we can understand
must consist of names which refer to sense-data with which we are acquain­
ted. A proposition is meaningful only if all of its real constituents stand for
something, and only the existence of sense- and memory-data is immune to
Cartesian doubt. 'This is white', referring to a present sense-datum, is about
'as simple a fact as one can get hold of, but Russell did not rule out that
the analysis of propositions might 'go on forever' (Logic 198-202).
The possibility of openended analysis was unacceptable to the early Witt­
genstein. His quasi-Kantian theory of symbolism left the actual 'composition'
of elementary propositions to the 'application of logic': only future analysis
could reveal the composition and logical forms of elementary propositions
(TLP 5.557). But he insisted 'on purely logical grounds' (TLP 5.5562) that
there must be elementary propositions to ensure that the analysis of proposi­
tions terminates, that the sense of propositions is determinate, that no truth-
value gaps should occur, and that whether a proposition has sense should
not depend on empirical facts. Elementary propositions form the basis of all
linguistic representation (see GENERAL PROPOSITIONAL FORM), and hence the core
of the PICTURE THEORY. Wittgenstein does not decide what propositions are
unanalysable, but specifies, more rigorously than Russell, the conditions they
have to fulfil. They must be:
(a) Logically independent. No two elementary propositions can either be
inconsistent with or entail each other. If '/»' entails 'q', its sense con­
tains that of '<?', that is, analysis must reveal 'q' to be one of p's truth-
102
ELEMENTARY PROPOSITION
functional components. Equally, if 'p' contradicts 'q\ it entails, and
hence 'contains', '~q'. In both cases 'p' is complex, not elementary
(TLP 4.1211, 4.211, 5.134, 6.3751). This requirement was fuelled by
the idea that molecular propositions are truth-functions of elementary
ones, which presupposes that in a TRUTH-TABLE each elementary propo­
sition can be assigned a truth-value independendy of all others.
Pictures. They depict a 'state of affairs', assert the existence of a cer­
tain combination of OBJECTS. If they are true, that state of affairs exists,
it is what Russell called an 'atomic fact' (TLP 4.21).
(c) A 'nexus' or 'combination' of names. A fully analysed proposition con­
sists exclusively of logically proper NAMES 'in immediate combination',
which go proxy for simple objects. Elementary propositions depict
states of affairs by combining the names in a way which corresponds
to a possible combination of objects (TLP 4.22f).
(d) Intrinsically positive. Condition (b) implies that all elementary proposi­
tions depict, truly or falsely, a 'positive fact', namely the existence of a
state of affairs. They say that something is the case, that objects are
combined in a certain way, rather than that, something is not the case
(TLP 4.021-4.023). And condition (a) implies that if '/>' is elementary,
l
~p' cannot be, since the two are contradictories. A false elementary
proposition is not the negation of a true one, but rather depicts a
different, and non-existent, combination of objects (TLP 2.06, 4.022;
RUL 19.8.19).
(e) Capable of being false in only one way. Propositions about complexes
(e.g., '<t>(aRby) can be false either if the complex does not exist (i.e., if
a does not stand in the relation R to b), or if it lacks the property
attributed to it (O). An elementary proposition, by contrast, excludes
exacdy one possibility, namely that the objects named by its constitu­
ents are not arranged the way the latter are in the proposition (TLP
4.25f.).
Some commentators hold that Wittgenstein was deliberately non-commit­
tal about any other features of elementary propositions, since they are in­
essential to the transcendental deduction of their existence. But Wittgenstein
inherited other ideas about the nature of propositions from Frege and Rus­
sell, notably that they are composed not of subject and predicate, but of
function and argument (TLP 3.141, 3.318, 5.47). Russell maintained that
the simples named by the constituents of atomic propositions comprise not
just 'particulars', but also 'qualities' like colours, and 'relations'. Wittgenstein
initially rejected this view, holding that a proposition like 'Socrates is
human' is not of the form Fa, but to be analysed as 'Socrates' and 'some­
thing is human', and that objects are not of different logical types (RUL
1.13; NL 100, 107). He abandoned the former claim (see GENERALITY), and
103
 ELEMENTARY PROPOSITION
his Notebooks state explicitly that 'relations and properties are objects too' (NB
16.6.15, see 21.6.15; NM 112).
Nominalist interpreters maintain that the Tractatus, by contrast with the
Notebooks, treats properties and relations as logical forms, not objects; elemen­
tary propositions are logical networks sprinkled with names of particulars.
They have adduced four arguments. The first is that the Tractatus indicates
that signs for properties and for individuals are of different logical types, and
employs different styles of variables for them (TUP 3.323f, 5.5261); Wittgen­
stein would not have failed to mention that there are two distinct types of
objects, consequendy the difference is one between names (which stand for
objects) and other signs. But through the claim that they have different LOGI­
CAL FORMS Wittgenstein did sort objects into distinct categories of different
combinatorial possibilities. He may have thought it superfluous to state
explicidy that the most general distinction is between individuals, properties
and relations, given Russell's similar position.
The second is that Tractatus 2.0251 states that 'Space, time, and colour
(being coloured) are forms of objects.' But what are here called forms of
objects are not determinate properties (spaces, times, colours), but determin-
ables like being coloured (see TLP 2.0131). Rather, that the Tractatus speaks
of such deteraiinables as 'formal properties', and also of 'formal relations',
suggests that there are also non-formal properties and relations (TLP 4.122).
Finally, since an object's form comprises its possibilities of combining with
other objects, Tractatus 2.0251 and Proto-Tractatus 2.025 If. imply that visual
objects combine with colours.
The third argument is that the comparison of propositions with spatial
arrangements (TLP 3.1431, 4.012, 4.016, 4.0311) suggests that in an ideal
notation properties and relations are displayed not through function-signs,
but through spatial properties of names standing for particulars: fa' is
expressed by 'H' and '(b(xj)' as 'y\ It has been objected that this would
imply, contrary to the Tractatus, that the depth-structure of propositions must
be expressed in writing. Yet the nominalist proposal is committed only to
the possibility of replacing function-signs by relations (spatial or temporal)
between signs. But it does ignore that neither the indefinite number of possi­
ble properties and relations, nor their different logical multiplicities, can be
displayed through discernible configurations. To avoid this difficulty, it has
been suggested that relation-signs occur in elementary propositions without
being names. But this contradicts condition (c): the only components of ele­
mentary propositions are names.
Lastiy, it is pointed out that the Tractatus holds that, 'Instead of "The
complex sign 'aRb' says that a stands to b in the relation R", we ought to
put, "That 'a' stands to 'b' in a certain relation says that aRb"' (TLP
3.143If). But this passage is directed not against the idea that relations are
objects but against Russell's claim that 'aRb' is the name of a complex in
T04
ELEMENTARY PROPOSITION
which both relata are in turn linked to the relation R According to Witt­
genstein, objects combine in states of affairs not with the aid of further links,
but direcdy, like links in a chain. The components of states of affairs stand
in a determinate relation to each other (aRb is not identical with bRa) with­
out any logical glue. The representation of this is possible because PROPO­
SITIONS are facts. What represents the relation between a and b in 'aRb' is
not 'R' as such, but that it occurs between 'a' and 'b'. The real component
of 'aRb' that signifies that relation is not 'R', which looks like the proper
names 'a' and 'b', but 'xRy', which is a relation-name (see NL 96-8; TLP
2.03f.).
That some names stand for properties and relations is further suggested
by three points. First, it is the only, way of reconciling two claims about ele­
mentary propositions: that they are a nexus of names - (c) - and that they
consist of function and argument. Secondly, according to Tractatus 4.24, ele­
mentary propositions are functions of names and have the form fx, (j>(x,y),
etc. Thirdly, Tractatus 4.123 speaks of shades of colours as objects, at least in
an extended use of the term. The realist interpretation is further supported
by Wittgenstein's subsequent discussions of the Tractatus. Most notably, in a
lecture Wittgenstein unequivocally stated that the* objects of Tractatus 2.01
include properties like colours and spatial relations (LWL 120; see RLF;
WVC 220; PG 199-201; TS22d §109; MS127 1.3.44). Moreover, he
ascribes to the Tractatus the view that a property is an object which can
enter into combinations with individuals (GB 134; BT 433-4; BB 17).
The Tractatus's failure to provide examples of elementary propositions is
due less to agnosticism than to the difficulties Wittgenstein encountered (in
the Notebooks) in trying to square his preconceptions about simplicity with his
logical specifications. Nevertheless, hints in the Tractatus, as well as in pre­
vious and later writings, indicate that analysis proceeds in the direction of
the phenomenally given (sensory impressions). States of affairs are instantia­
tions or co-instantiations of properties like colours and (spatial) relations at
spatio-temporal points or points in the visual field. A point in the visual field
stands in a 'colour-space': it must have some colour, and combines with a
particular colour, like two links in a chain without any additional relations
(TLP 2.0131; NB 3.9.14, 6.-7.5.15; PG 211). 'Some Remarks on Logical
Form' makes the picture concrete: take your visual field, flatten it and put a
grid across. Elementary propositions use the coordinates to refer to a point
in the visual field, and ascribe to it a shade of colour (a system reminiscent
of PI §48; see also ROC I 61-2, III 58, 149), for example,
(1) A (the spatial point with coordinates x,y) is red.
Accordingly, objects are minima sensibilia (NB 7.5.15): particulars like spatial
points, ultimate perceptual qualities such as shades of colours, tones and
smells and simple spatial relations.
105
ELEMENTARY PROPOSITION
One objection against this interpretation is that the OBJECTS of the Tracta­
tus must be indestructible, common to all possible words. But unlike Rus­
sell's sense-data, Wittgenstein's candidates are not fleeting mental episodes.
They are not temporary, and might appear to be sempitemalia the existence
of which is metaphysically, not just epistemically, guaranteed. Red com­
plexes and sense-data can be destroyed, but, according to the position Inves­
tigations §§46-59 ascribes to the Tractatus, the colour red cannot. The same
might be thought of spatio-temporal points: while such a point might fail to
have a certain colour, it could not fail to exist. That Wittgenstein extends
sempiternality to points in the visual field can be explained by reference to
his SOLIPSISM, which insists that the world is what is given to a transcendental
subject of representation.
A second objection is that Tractatus 6.3751 states that a proposition like (1)
is incompatible with, for example, 'A is green' and hence not elementary.
Wittgenstein thought that (1) can be analysed as a logical product of ele­
mentary propositions which entails that A is not green, and seems to have
envisaged that the resulting elementary propositions ascribe to A either
unanalysable shades of colour, or light of a certain wavelength. In 1929 he
realized that this programme is hopeless. The resulting propositions again
exclude each other (if A is dark red it cannot be light red, if it only emits
light of 620 nm it cannot also emit light of 520 nm). The reason is that like
(1) they ascribe to an object one out of a range of incompatible specifica­
tions, a determinate of a determinable colour, velocity, electrical charge,
pressure, etc. And there is no way of analysing such propositions into simple
ones which would satisfy the requirement of logical independence (see
COLOUR). Wittgenstein's reaction was to abandon not the idea that elemen­
tary propositions involve phenomenal qualities, but the insistence on the
logical independence of elementary propositions, and with it logical atomism
(RLF; PR ch.VIII; MS 105) (Russell had always been less rigorous in this
respect, and hence less troubled by colour-exclusion). Elementary proposi­
tions may exclude each other. What is compared with reality is never a
single proposition, but a 'proposition system': (1) at a stroke determines that
A is neither green nor blue, etc. (WVC 63-4; PR 109-12).
In any event, Wittgenstein came to believe that nothing could satisfy his
specifications for elementary propositions. Take another candidate with
which he had toyed (NB 29.10.14, 20.6.15), propositions which ascribe a
spatio-temporal location to physical simples.
(2) The material point P is in place xj,z at time /
excludes P's being in any other place, and hence again is not elementary. It
has been suggested that the problem is avoided by propositions which
simply combine spatial and temporal coordinates:
106
ETHICS
(2') xy,z,t.
This suggestion honours condition (a), since (2') implies nothing about other
spatio-temporal points. However, it violates (b). (2') by itself is a merely the
name of a point. To turn it into a picture of a state of affairs which states
that a mass-point exists at a certain spatio-temporal point, one has to add a
quantified provision: 'There is a mass-point...' That is, (2') itself is not a
proposition. Without reference to qualities and relations nothing can be said,
and most qualities and relations are determinates of a determinable. Conse-
quendy, even if one can construct logically independent propositions, it is
improbable that one can analyse ordinary propositions into such proposi­
tions.
With the demise of logical atomism, elementary propositions lose their
'earlier significance' (PR 111). However, the notion soldiers on for a while
with the idea that the only genuine propositions are sense-datum propositions
that describe immediate experience. This position is closer to Russell than to
the Tractatus, in that it makes semantic prirnitiveness turn on epistemic prirni-
tiveness, and it influenced logical positivism's idea of an observation-sentence.
Wittgenstein later rejected it (see PRIVATE LANGUAGE ARGUMENT). He also
claimed that propositions are simple only in the relative sense that within a
given grammatical system there are no provisions for their truth-functional
analysis (PG 211), as with the colour-propositions of Investigations §48.
ethics Ethics occupies a peculiar role in Wittgenstein's thinking. He
attached overarching personal importance to questions of moral value. Yet,
his written treatments of ethics are brief and' obscure, while his views on
language have had a strong, albeit intermittent and diffuse, influence on
analytic moral philosophy. Wittgenstein's personal moral oudook was ego­
centric and contemplative. In this he was shaped by.Schopenhauer and by
Weininger's Sex and Character, which proclaimed that 'logic and ethics are
fundamentally the same, they are no more than duty to oneselF (159). One
has a moral obligation to strive for logical clarity. The SAYING/SHOWING dis­
tinction of the Tractatus gives substance to the first part of Weininger's
slogan: only the empirical propositions of science are meaningful, since they
picture contingent states of affairs (truly or falsely). What Wittgenstein calls
the 'higher' (TLP 6.42, 6.432), all areas of value, share with the logical
structures of language the fate of being ineffable; they cannot be said, they
can only be shown. Ethics, aesthetics and logic are linked by virtue of being
'transcendental': while everything factual is 'accidental', they try to express
what could not be otherwise, the 'preconditions of the world' (NB 24.7.16;
TLP 6.13, 6.421).
However, unlike the logical structure of language, ethical value is not
even shown by any meaningful propositions, although it may be shown in
107
 ETHICS
actions, attitudes or works of art (EL 9.4.17). Ethics is not just transcenden­
tal, it is 'transcendent'. Values 'cannot lie within the world', which 'itself is
neither good nor evil'; their 'bearer' is a Schopenhauerian metaphysical WILL
outside the world (TLP 6.41-6.43; NB 2.8.16). Wittgenstein resolves an
inconsistency between two Schopenhauerian ideas, namely that moral
redemption lies in denying the will and that compassion, an exercise of the
will, is essential to morality, by adopting a Kantian distinction between good
and bad willing (NB 21./24./29.7.16; TLP 6.43; World II chs XLVII-
XLLX). Equally Kantian is the view that the consequences of an action are
ethically irrelevant, unlike the spirit in which it is performed. But Wittgen­
stein's rationale is Spinozistic rather than Kantian. He identifies being good
with being happy, being bad with being unhappy (NB 8./29./30.7.16).
Reward and punishment are crucial to ethics, but 'reside in the action itself
(TLP 6.422). The reason is that the will is causally impotent. Good or evil
willing cannot change the facts, but only the 'limits of the world', namely
the 'attitude of the subject to the world'. A good will is its own reward
because it looks at the world with 'a happy eye', accepts whatever happens
with equanimity (TLP 6.43; NB 20.10.16). This Stoic attitude is the ethical
result of the MYSTICAL capacity to view the world sub specie aetemitatis, which is
also essential to art. 'Ethics and aesthetics are one' not just because they are
ineffable, which is merely a precondition for their identity, but because both
are based on a mystical attitude which marvels at the existence of the world,
and is content with its brute facts (TLP 6.421, 6.45; NB 7./8.10.16).
Wittgenstein maintained that the 'point' of Tractatus Logico-Philosophicus was
an ethical one, namely to delimit 'the Ethical' from within, by 'remaining
silent about it' (FL 10./11.19). However, the structure and composition of
the Tractatus suggest rather that the mystical passages owe their existence to
Wittgenstein's experiences during the war, and were grafted onto a logical
trunk (the connection being provided by the saying/showing distinction).
This is confirmed by the fact that ethics plays only a minor role in Wittgen­
stein's subsequent rethinking of the Tractatus. Most notable is 'A Lecture on
Ethics' of 1929, which elaborates the idea that ethics is ineffable. It expands
Moore's definition of ethics as the enquiry into what is good, in order to
accommodate everything that has value and concerns the meaning of life,
including aesthetics. Again following Moore, Wittgenstein distinguishes a tri­
vial or relative, from an ethical or absolute sense of terms of appreciation. The
relative sense simply implies satisfaction of certain standards, as in 'You play
tennis well.' By contrast, the absolute sense is elusive, since no factual state­
ment can ever be or logically imply an absolute judgement of value such as
'You ought to behave decendy.' Wittgenstein invokes three experiences to
shed light on absolute value. The first is the mystical experience of wonder
at the existence of the world. The second is the feeling 'I am safe, nothing
can injure me, whatever happens.' This Stoic thought is notorious from
108
ETHICS
Socrates and Kierkegaard. In Wittgenstein's thought it follows from the logi­
cal independence of world and ethical will: just as the latter cannot influ­
ence the former, the world cannot harm a virtuous person. For goodness
lies in the eye of the beholder, in meeting the afflictions of life in a happy
spirit. In this sense, the world of the happy, that is, virtuous, person differs
from that of the unhappy one (NB 29.7.16; TLP 6.43). The final experience
is that of guilt, which Wittgenstein explains as God's disapproving of one's
conduct. In the same vein, he rejects the 'rationalist' claim that 'God wills
the good because it is good' in favour of 'Good is what God demands',
on the grounds that this reveals the inexplicability of the good, and its
independence from facts (WVC 115).
These three points deliberately explain the obscure - absolute value - by
the more obscure. The last one makes ethics parasitic on RELIGION, only in
order to insist that ethics cannot be explained. Moreover, it falls foul of an
elenctic argument in Plato's Euthyphro which is close to Wittgenstein's own
procedure: we wouldn't call murder 'good', even if it were demanded by
God. And Wittgenstein himself acknowledged that the first two amount to a
misuse of expressions like 'safe', 'existence' or 'wondering'. Making a virtue
of necessity, he radicalizes Moore's claim that 'good* is indefinable: ethics is
deep precisely because it inevitably transgresses the limits of language. For­
tunately, this is wrong: while judgerrfents of absolute value like 'One ought
to keep promises' may not be factual, they are neither nonsensical nor mys­
tical in the way Wittgenstein envisages. Indeed, his insistence on the inef­
fable nature of ethics is explicidy stipulative: 'I would reject any significant
description [of ethics] ah initio, on the ground of its significance.' Behind this
stipulation lies the conviction that language can express only facts, which
restricts significance to factual description (LE 7-9, 11-12; WVC 68-9).
This credo, part of the PICTURE THEORY, is later abandoned. It may be
'impossible' to describe what ethical (and aesthetic) appreciation consists in,
but the reason lies in CONTEXTUALISM: we must focus not on the appearance
of ethical terms, which resembles that of other words, but on their specific
role within our whole culture (LC 2, 7-8). The ethical shows itself no longer
in mystical attitudes of a solipsistic self, but in social patterns of action. As a
result, sibylline pronouncements on the ^definability or ineffability of ethical
terms give way to (underdeveloped) investigations into their use (RPP I
§160; AWL 34-6). One result of this investigation is that the 'meaning of
"good" is bound up with the act it modifies' (a good lie is different from a
good deed). Wittgenstein concludes that 'good' is a FAMILY-RESEMBLANCE term.
But his argument establishes only that 'good' is used attributively rather
than predicatively (a good liar is not necessarily good tout court).
It is doubtful whether different things are termed 'good' because of over­
lapping similarities. That 'good' has a single ethical role in spite of different
standards of application is suggested by ideas which survive from the transi-
109
 ETHICS
tion period. Crucial to ethics is its contrast with factual propositions and
scientific theories. In spite of itself, 'A Lecture on Ethics' does explain that
contrast, at least partially, namely by reference to the action-guiding nature
of ethical judgements. While there is a logical gap between factual judge­
ments and decisions to act, ethical judgements express direcdy the grounds
or attitudes on which we act. This ties in with two later observations (LC 2;
AWL 35). Firsdy, ethical terms replace and extend natural reactions (ges­
tures, facial expressions) of approval and disapproval. Secondly, their 'gram­
mar' is determined not only by the object they modify, but also by the
reasons a person offers for applying them.
Wittgenstein draws relativistic conclusions from these observations (SDE
23-4; see FORM OF LIFE). Ethical judgements are not responsible to reality, and
do not contradict each other in the way empirical propositions do. They
express the reasons on which we act, and can be justified only within an
ethical system, such as Christian ethics. Like grammar, these systems are
AUTONOMOUS. Each one of them sets its own standards of justification, since
it involves a distinct array of moral concepts. Christian and secular ethics
employ terms like 'good' with different meanings, which means that their
claims are mutually incommensurable. This is not to say that divergent
judgements are 'equally right', or 'right from their own standpoint', but only
to say that to make them is to 'adopt' a certain framework of action
and justification, which itself cannot be justified. The question of whether
Christian or secular ethics itself is right 'does not make sense'.
Wittgenstein mentions one problem with his relativism, the thought that it
might destroy the 'imperative in morality'. He can allow for commitment in
the 'first person' (SDE 23; WVC 116-17; CV 60), since ethical judgements
express the basis on which an individual intends to act. However, he cannot
allow for the idea of moral obligations that bind all individuals, indepen-
dendy of their personal oudook. We can condemn actions demanded by
other ethics only from within our own system. If it is impossible to establish
the moral superiority of that system, we lack any justification for interfering
with such acts, although this is precisely what a universal obligation would
bid us do. Wittgenstein's discussion of the laws of logic suggests a strategy
for alleviating this problem, namely that there are conceptual limits on what
we call 'an ethics'. Unfortunately, this runs counter to his claim that even
Goering's remark 'Right is whatever we like' expresses 'a kind of ethics'
(SDE 25). But the fact that this slogan expresses the basis on which Goering
acted is not enough to sustain Wittgenstein's assessment. The point is not
that Goering's oudook is unintelligible, as some Wittgensteinians have sug­
gested, but that it is a paradigm of immorality, not an alternative ethics.
The logical positivists took over Wittgenstein's early claim that ethics is
nonsensical, because unverifiable, while dropping the idea of its paramount
importance. His later stress on non-descriptive uses of language influenced
110
EXPLANATION
both emotivism and prescriptivism. Ironically, he has also been invoked by
contemporary cognitivists, who claim that all indicative sentences, including
moral ones, make claims to truth. This suggestion is at odds with his view
that the similarity in linguistic form disguises logical differences between
moral and descriptive propositions (see SDE 24). But he shares the cogniti­
vists' idea that moral discourse cannot be disqualified as less objective than
scientific discourse.
explanation Although the Tractatus drew a sharp contrast between PHILO­
SOPHY and empirical SCIENCE (NL 106; TLP 4.11 Iff.), it can be seen as pro­
viding quasi-scientific explanations. Just as science explains the behaviour of
macroscopic objects by reference to their microscopic structure, so the Trac­
tatus explains the ability of ordinary language to depict reality by reference
to its hidden LOGICAL SYNTAX. This fact lies behind Wittgenstein's later exhor­
tations that philosophical explanations be replaced by descriptions of gram­
matical rules (PI §§109, 126, 496). Attempts to explain the INTENTIONAL
relation between language and reality by reference to logico-metaphysical or
mental mechanisms are spurious. Genuine CAUSAL? explanations are of course
legitimate, but their place is in the hypothetico-deductive sciences.
Wittgenstein's philosophy aims at a kind of understanding, yet one which
does not require discovering new evidence or hidden causal processes, but is
achieved by an OVERVIEW which organizes familiar phenomena in a new
way. One kind of explanation Wittgenstein provides pinpoints the sources of
philosophical confusions, but he intimates that unlike the diagnostic explana­
tions of medicine, such explanations are not causal. He detected various
similarities between his philosophical therapy and psychoanalysis: (a) both
try to bring out a patient's repressed worries; (b) the ultimate standard for
articulating these worries is that the patient should recognize them; (c) they
involve a fight against the will as well as the intellect; (d) the disease can
only be cured after it has run its course (AWL 37-40; PI §§133, 254-5, 599;
BT 407-10; PG 382; LC 18, 23-5, 43; Z §382).
A far more important role in his later work is played by explanations of
meaning. These are not causal explanations of why we use a certain term, or
of what the (perlocutionary) effects of using it are on particular hearers, but
explanations of how we use it, that is, they specify rules for its correct use
(PI §§120, 491-8). Such grammatical explanations are not, therefore, incom­
patible with the idea that philosophy describes linguistic rules. Unlike causal
explanations, which in principle can go on for ever, such explanations come
to an end. We cannot explain (save perhaps in the causal sense), for exam­
ple, why locutions like 'I was going to <X>' need not be based on evidence. It
is a characteristic philosophical mistake to look for a further explanation
here, when 'we ought to look at what happens as a "proto-phenomenon"'
111
 EXPLANATION
and simply to note 'this language-game is played' (PI §§654—5; Z §§314-15;
RFM 102-3; RPP I §889).
Wittgenstein claims that it is fruitful to investigate how a word is taught.
But this is not because he is engaging in armchair learning-theory (LC 1-2;
Z §412). Even his claim that teaching through explanations presupposes cer­
tain fundamental linguistic skills is not an empirical genetic theory, but con­
ceptual: explanations are correlates of requests for explanations of, or of
unclarities about, meaning, and hence presuppose a certain degree of lin­
guistic understanding on the part of the learner, for example the ability to
ask for the meaning of a word (PI §§6, 27; PG 62; PLP 126). Wittgenstein's
only contingent observation is that we are not born with such abilities, but
acquire them through 'training' (Abrichtung) or 'drill'. He also makes a peda­
gogic claim which is reminiscent of his observation that even doubt pre­
supposes the recognition of some authority: educators should keep in mind
that training provides the foundation of explanation, as it does for rule-
following or calculation (Z §419; PI §§5, 86, 157-8, 189, 198, 206, 441;
LFM 58—60, 182—8; see FRAMEWORK). Training does not presuppose under­
standing, but only patterns of reaction on the part of the trainee. A child
will look in the direction in which one points, while a cat will look at the
pointing finger. Wittgenstein also claims that the order of teaching is a
necessary condition for any logical priority between concepts: 'seems F
cannot be logically prior to 'is F if it must be taught later (PI §§143-6;
AWL 102; Z §§414-15).
Historical and physiological facts about how we are taught to speak are
philosophically irrelevant, what matters is what is taught (LWL 38; BB 12-
14; PG 41, 66, 70). In teaching, what we explain is the meaning of words.
An explanation of meaning, unlike mere drill or a drug which induces
understanding, is normative, it provides a standard for the correct use of a
term. In this respect explanations are linguistic rules, a point which makes
plausible Wittgenstein's idea that language is structured by GRAMMAR, a
system of rules (PG 191; TS228 §34). Equally important are the con­
sequences for Wittgenstein's elucidation of the notion of meaning. Meaning
is what is explained by an explanation of meaning. This allows one to side­
step the misbegotten search for the meaning of a term 'X', an entity of some
kind (a MEANING-BODY), in favour of an investigation of the way 'X' is
explained (PG 59, 69; BB 1; PI §560; AWL 48-9), and emphasizes the
normative nature of meaning: what explanations of meaning explain is the
correct USE of 'X'.
It also links explanation with linguistic understanding. The meaning of 'X'
is what one understands when one understands its explanation (BT 11; PG
60). Both explanation and use are criteria of UNDERSTANDING a word.. To
understand 'X' is not only to be able to use it correcdy, but to be able to
answer the question 'What does "X" mean?' Wittgenstein's remarks on the
112
EXPLANATION
conceptual connections between meaning, explanation and understanding
sound trivial, as they should, being grammatical reminders. But if correct,
they have significant philosophical implications. For one thing, they imply
that neither the meaning of a word, nor our understanding of it, can out­
strip our ability to explain it (PI §75). A speaker may understand more than
he explains, but not more than he is able to explain. This would rule out
the CALCULUS MODEL, according to which the meanings of words and our
understanding of them are determined by hidden rules of which we are
ignorant.
However, the claim that understanding parallels explanation must itself be
qualified. Thus, Wittgenstein recognized that in exceptional circumstances
the two criteria of understanding can come apart: someone may be able to
apply 'X' correcdy, without being able to explain it at all. Moreover, it is
quite common that we can only give defective explanations for certain
terms, notably those which have to be explained for different contexts, like
prepositions or conjunctions. This is not always a sign of carelessness or lack
of linguistic self-consciousness: a satisfactory explanation of such terms
requires thorough elicitation and reflection. Wittgenstein tended to ignore
this point; however, it does not follow that the correct explanations can be
such that competent speakers uncoached by philosophers would not even
recognize them, as is the case with the logicists' definition of numbers as sets
of sets and many explanations given by contemporary theories of meaning.
Frege's idea that we can discover that the real meaning of a word differs
radically from the meaning we have given it in our explanations is inco­
herent (BB 27—18; Foundations vii).
The philosophical tradition is inimical to this suggestion. Since Plato it
has been assumed that the only adequate or legitimate explanation of a
term is an analytic definition, which analyses it into a conjunction of char­
acteristic marks, preferably per genus et differentiam. Thus, Frege treats defini­
tion as logical analysis into 'marks' (Merkmale) which together make up the
definiendum. He relaxed the initial requirement by allowing for definitions
which use expressions of generality instead of marks (e.g., in case of 'prime
number'), but insisted that definitions must specify necessary and sufficient
conditions for the application of a word (Foundations §§53, 104; 'Concept').
For Russell, definitions are symbolic abbreviations constructed out of
primitive (i.e., unanalysable) ideas (Principles 27, 429; Principia i.l, 91), and a
similar idea is at work in the Tractatus's conception of LOGICAL ANALYSIS.
Plato also suggested that we cannot find out anything about X unless we
possess an analytic definition of X. Accordingly, such definitions must stand
at the beginning of a philosophical system, an idea embraced by the ration­
alists. Kant demurred, because he felt that stringent definitions can at best
be the result, not the starting-point, of philosophical inquiry. But the con­
ception of what constitutes a philosophically adequate explanation remained
113
EXPLANATION
unchallenged until Wittgenstein's later work: 'I cannot characterize my
standpoint better than by saying that it is opposed to that which Socrates
represents in the Platonic dialogues' (TS302 14; PG 120—1). Socrates was
right to ask, for example, 'What is virtue?', but wrong to reject partial
explanations or explanations by exemplification or analogies. There may be
reasons for restricting 'definition' to explanations of a certain kind. But it is
wrong to hold that the meaning of 'X' or the content of our understanding
of it is equivalent to such a definition (PI §75).
Forms of explanation are diverse, analytic definition is only one of them.
Other legitimate forms of explanation are. OSTENSIVE DEFINITION, paraphrase,
contrastive paraphrase, exemplification, series of examples, etc. These ordi­
narily accepted explanations are not defective or incomplete. Examples, in
particular, 'are decent signs, not just rubbish or hocus-pocus' (PG 273). Not
only are some terms inaccessible to analytic definition, notably coLOUR-terms
and FAMILY-RESEMBLANCE terms; the idea that philosophically adequate expla­
nations must be complete in the way that such definitions are complete is
misguided to begin with. The function of such explanations is to remove or
avert misunderstandings which do or would actually occur without them,
not all conceivable misunderstandings (PI §88). This does not mean that a
correct explanation is simply one which actually results in understanding.
However, it does mean that a correct explanation of 'X^ need not cover all
the circumstances in which it can be used, but only discriminate relevant
circumstances in which 'X' can and cannot be used. The concept of com­
pleteness is purpose and circumstance relative. An explanation is complete if
it can be invoked as a standard for the correct application of a term in
normal contexts. An explanation of 'thinking' need not predetermine whe­
ther or not fish think, an explanation of number (pace Frege) need not tell
us that Caesar is not a number (Z §§114-18; BT 60-9; Foundations §56).
Wittgenstein here conflates what an adequate explanation must include
and what it must convey. An explanation of 'thinking' need not mention
fish or imply anything about them, but it must indicate possible grounds for
deciding whether or not they think. Equally, an explanation of number need
not mention Caesar, but must indicate the category difference between
numerals and the names of people. At the same time, Wittgenstein is right
to insist that no explanation can forestall the mere possibility of mis­
understanding or doubt (PI §§80, 84-7). An explanation is adequate if
it establishes an agreed pattern of application relative to a certain set of
FRAMEWORK conditions. For example, our criteria of personal identity com­
bine bodily continuity, memory and character-traits. If these no longer coin­
cided, the term 'person' would disintegrate. But that logical possibility does
not render our present explanation of 'person' inadequate (BB 62).
114
F
fact 'The world is everything that is the case. The world is the totality of
facts, not of objects.' The famous beginning of the Tractatus is the climax of
a realist tradition which assigned importance to facts as mind-independent
constituents of the world. Frege, Moore, Russell, and Wittgenstein in 1911
combined this motif with a (partly terminological) idiosyncracy: they identi­
fied facts with 'true' or 'asserted propositions'. But Russell and Wittgenstein
soon came to understand facts as what makes propositions true (if they are
true). Like Moore, Russell treated a fact as a complex of entities ('concepts'
or 'terms') which subsists tunelessly, regardless ofwhether it is thought by
anyone: the fact that Socrates is mortal consists of the philosopher and the
property of being mortal. In his atomistic period, he analysed the world into
'atomic facts' consisting of simple 'individuals', which comprise 'particulars',
their qualities and relations (Principles ch. 4; Logic 178-89; Writings ch. 1).
At first, Wittgenstein maintained that the MEANING of a proposition p is the
fact that corresponds to it in reality, the fact that p if it is true, the fact that
~p if it is false. Later, he abandoned this idea. Only NAMES have a meaning,
the absolutely simple 'objects' they stand for. Propositions do not, since they
do not stand for anything, but describe; and what a proposition describes, a
fact or situation, can be expressed only by a proposition, something which
can be prefixed by a that-clause. In spite of occasional lapses (NB 6.10.14,
30.5.15), Wittgenstein urged against Moore and Russell that neither a pro­
position nor what it represents is a 'complex' (TLP 3.14ff.; NL 98, 107).
Complexes are mere combinations of objects, and are denoted by definite
descriptions; they include what we ordinarily think of as objects (TLP
2.0201, 3.24; NB 23.-24.5./15.6.15). Like complexes, but unlike OBJECTS,
facts are composite, composed of simples (aus einfachen Teilen zusammengesetzt;
NB 17.6.15; see TLP 3.21, 4.032). Propositions are themselves facts, not
mere lists of names: what symbolizes in 'aRb' is not the complex of signs,
but the fact that 'R' occurs between V and 'b' with 'a' to the left and T to
the right. By the same token, facts in general are distinct from complexes of
objects: the fact that a stands in the relation R to b is distinct from the com­
plex (aRb) - a's standing-in-the-relation-i? to b - which is itself a constituent
of a fact like Q>(aRb). The broom consists of the broomstick fixed in the
115
 FACT
brush, but it is a component of facts - e.g., the fact that the broom stands
in the corner - rather than itself being a fact. A fact or state of affairs
cannot be identified by listing its components, but only by specifying the
determinate way in which these are connected, its 'structure'. Whereas the
complex (aRb) is the same as the complex (bRa), the fact that aRb differs
from the fact that bRa; looking at the Necker cube we can perceive two dis­
tinct facts with the same constituents (TLP 2.032, 3.1432, 5.5423).
Russell sometimes follows Bradley in holding that the components of the
relational fact aRb need to be bound together by further relations relating a
and b respectively to R; at other times he maintains that what unites them is
a logical form - x<S>y ('Theory' 80-8, 97-9). Wittgenstein's distinction of
facts and complexes renders both suggestions superfluous. The LOGICAL FORM
of a fact is not one of its constituents. aRb and bRa have the same con­
stituents but are different facts. A two-place relation needs only two mona­
dic objects - a and b — and a diadic one - xRy — to constitute a state of
affairs, not two further relations to link each object with the relation
between them. In a state of affairs, objects fit into one another, like links in
a chain, without any logical glue.
According to the PICTURE THEORY, a proposition represents its sense, a state
of affairs, which may or may not obtain, depending on whether the proposi­
tion is true or false (TLP 2.20Iff., 4.02If., 4.031). There is a terminological
unclarity here. 'What is the case, the fact, is the existence (Bestehen) of states
of affairs. A state of affairs is a combination (Verbindimg) of objects' (TLP 2f.;
note that the literal translation of Bestehen is 'obtaining', and that single-
object states of affairs are excluded by definition). In a letter to Russell,
Wittgenstein stated that a state of affairs (Sachverhalt) is what corresponds to
a true elementary proposition (e.g., '/>'), while a fact (Tatsache) is what corre­
sponds to a true molecular proposition (e.g., 'p.q.r') (RUL 19.8.19); and he
approved of Ogden's corresponding translation of Sachverhalt as 'atomic fact'.
Nevertheless, 'state of affairs' is the literal translation, and does not beg exe-
getical questions. For there is also evidence that the difference between
states of affairs and facts is the difference between what is possibly and what
is actually the case. The Tractatus applies the terms 'possible' and 'non-
obtaining' to states of affairs and situations (Sachlagen), but never to facts
(TLP 2.012ff., 2.06, 2.202f, 3.11). At the same time, facts are more complex
than states of affairs (TLP 2.03ff, 4.2211): a fact (its structure) consists of a
plurality of states of affairs (their structures). Therefore the following distinc­
tion has,been suggested: a fact is the existence of a set of states of affairs
(S\...S ); a state of affairs is a possible combination (concatenation/
n
arrangement) of objects corresponding to an elementary proposition; a situa­
tion is a possible arrangement corresponding to a molecular proposition.
However, while some passages support the suggestion that situations are the
molecular equivalent of states of affairs, others belie it (TLP 2.11, 2.20If. vs.
116
FACT
2.012ff., 2.034). In view of Wittgenstein's own pronouncements, one should
therefore treat the distinction between states of affairs and facts as one
between the elementary and the complex. Alas, this would render his posi­
tion inconsistent. The sense of a proposition, what it depicts, is a state of
affairs or situation (TLP 2.201ff, 4.02ff; NB 2.10./2.11.14). A state of
affairs is a possible combination of objects which exists if the proposition is
true, and does not if it is false, otherwise the sense of a proposition would
depend on its being true. On the other hand, to speak of possible or non-
obtaining facts runs counter to ordinary usage. By itself this would not
matter, since Philosophical Grammar explicidy rejects this terminological restric­
tion (PG 301-3; see also FW 55). But it is also incompatible with the Tracta­
tus's own statement that a fact is something which is the case (TLP Iff.).
Wittgenstein's letter ignores that states of affairs must be potentialities, facts
actualities.
Perhaps this is due to his operating with two different concepts of a fact.
Initially, a fact is the obtaining of a state of affairs. But the Tractatus also dis­
tinguishes between a 'positive fact' as the obtaining and a 'negative fact' as
the non-obtaining of states of affairs (TLP 2.06; NL 97-9). However, this does
not remove the aforementioned difficulty, since $ negative fact the fact
that such-and-such is not the case - like a positive fact, is an actuality.
Worse, there is an additional unclarity. The Tractatus defines the world as
the totality of obtaining states of affairs, reality as the obtaining and non-
obtaining of states of affairs, but also claims 'The sum-total of reality is the
world' (TLP 2.04, 2.06, 2.063). Together, these passages seem to identify the
set of positive facts with the set of positive and negative facts. One might try
to resist that conclusion by pointing out that the world is identified with the
totality of obtaining states of affairs, while reality could be a subset of the
totality of obtaining and non-obtaining states of affairs. But even this subset
must include non-obtaining states of affairs, which are not part of the world.
At any rate, the structure of reality is implicated in that of the world
(TLP 1.12, 2.05). Objects cannot occur on their own, but must enter into
combinations with other objects. A list of all positive facts therefore men­
tions all objects. Moreover, objects have not only external properties (of
being actually combined with other objects in facts), but also internal prop­
erties, the capacity of being combined with other objects in possible states of
affairs. Every object contains within its nature all the possibilities for its
entering into combination with other objects. This means that the totality of
objects, which is given with the totality of obtaining states of affairs (=
world), determines the totality of possible states of affairs (= reality). Indeed,
if even a single object a is given, all objects are given (TLP 2.011-2.014,
5.524). For the nature of a will determine for all other objects whether or
not they can be combined with a.
Although the Tractatus distinguishes between positive and negative facts,
117
 FACT
the facts that ELEMENTARY PROPOSITIONS depict the existence of are always
positive facts, namely the obtaining of a state of affairs (RUL 19.8.19). By
combining names in a certain way, an elementary proposition says, truly or
falsely, that something is the case, that objects are combined in a certain
way, rather than that something is not the case (TLP 4.022). Consequentiy,
the negation of a true elementary proposition is not a false elementary pro­
position, but a false molecular proposition, while a false elementary proposi­
tion depicts a non-obtaining combination of objects.
This approach solves a problem about negative facts and propositions. It
seems that what corresponds to the negative proposition 'Blood is not green'
must be blood's not being green, which in turn consists in its being red or
blue or yellow. Along such lines, Schlick reasoned that negative propositions
are essentially ambiguous and hence defective. Russell, by contrast, righdy
rejected the idea of analysing '~Fa' as 'There is a proposition "Ga" which is
true and incompatible with "Fa"', and hence grudgingly accepted negative
atomic facts as part of the ontological zoo (Logic 209-14). Wittgenstein avoi­
ded Schlick's confusion by distinguishing between />'s not being the case and
what is the case instead of p: 'not-/)' does not mean the same as 'anything
else, only not />'. But he also insisted that '~p' does not refer to a different
reality from '/»': it is the same fact which verifies one of them and falsifies
the other (NL 94-6; NB 4.11.14; TLP 4.062If.).
Wittgenstein's ontology of atomic facts can be illustrated as follows: a-d
are objects of one type (individuals), E-H of another (properties), shaded
areas are actual combinations (obtaining states of affairs), unshaded areas
possible but non-obtaining combinations:
a b c d
E
F
G
V,
Just as elementary propositions are logically independent, so are the atomic
facts and states of affairs they depict. The obtaining or non-obtaining of a
state of affairs must neither preclude nor entail the obtaining or non-obtain­
ing of any other (TLP 1.2f, 2.061f; NB 28.11.16) - F s being combined
with b neither precludes nor entails F s being combined with a or b's being
118
FACT
combined with H. Accordingly, what corresponds to the negative fact that
~Fa is not that objects a and F could not be combined because of G and a's
being combined, but simply that they are not so combined. A complete
description of the 'world', of the totality of positive facts, is given by listing
all elementary propositions and specifying which of them are true (TLP
4.063, 4.26). No two members of this list will be logically incompatible, and
none will be the negation of any other. We do not require any propositions
involving the negation-sign. One and the same reality corresponds to a pro­
position and its negation, and that reality is always a concatenation of
objects, that is, a positive fact.
This model underlies the idea that the world is the totality of facts rather
than things. Some have understood this as a novel, dynamic ontology, based
on the idea that we perceive the world not as composed of disconnected
bits, but as being ordered according to that-clauses. But the Tractatus is not
concerned with how we perceive reality. Its ontology is part of a theory of
symbolic representation. The world is primarily what is being represented in
language. That the world is the totality of facts means that in order to
represent the world we have to depict facts. In that sense, the world cannot
consist of, that is, be identified with, objects, since the latter are common to
all possible worlds.
Although Wittgenstein's later reflections on facts rarely engage with the
details of the Tractatus, they suggest that this ontology of representation is
based on a misconception of facts (PG 199-201; PI §48; MS 127 1.3.44).
Wittgenstein reiterates and elaborates the distinction between facts and com­
plexes. A complex (e.g., a flower) is a spatial object composed of spatial
parts smaller than the whole (e.g., stalk and bloom). The spatial relations
between the component parts are not themselves components of the com­
plex: a chain is composed of its links, not of its links plus their spatial rela­
tions. The same goes for the properties of a complex: a red circle consists of
parts, but it does not consist of redness and circularity. These observations
may be directed against the Tractatus, although it is unclear whether the
latter is committed to treating redness as a component of a complex. On
the other hand, Wittgenstein clearly criticizes the Tractatus for speaking of
facts and states of affairs as composed of constituents (Bestandteile), as 'combi­
nations' or 'configurations' of objects (TLP 2.01f, 2.0271-2.03). The fact
that the circle is red is not a combination of a circle and redness, the fact
that the book is on the table is not composed of the book, the table and the
relation of being on. Consequentiy, facts are not formed by objects as a
chain is by its links. Unlike complexes, facts are not spatio-temporal occu­
pants of the world (they have no spatial location and cannot move). For this
reason it is also misleading to conceive of propositions as describing facts,
states of affairs or situations (cp. TLP 3.144, 4.016, 4.023). Rather, in stating
a fact, for example that the broomstick is stuck into the brush, one can
119
 FAMILY RESEMBLANCE
describe a complex, namely the broom, which is a space-occupant. Similarly,
one cannot point at a fact, but only point out a fact. One can point at a
complex, but that is not to point out that its components are related in a
certain way. The Tractatus wrongly assimilated facts to constituents of the
world. The world is the totality of things, not of facts, although a description
of the world consists of statements of fact. This undermines not only logical
atomism, but also any correspondence theory of TRUTH, such as the one of
the Tractatus, which treats facts as worldly items to which our propositions
correspond.
family resemblance (FarnUumdhnUchkeit) The term occurs in Nietzsche
(Beyond Good and Evil §20). Another possible source is Nicod's discussion of
various types of resemblances (Geometry in the Sensible World 55fT.). Wittgen­
stein first uses it in 'Big Typescript' §58, which reprimands Spengler for
sorting cultural epochs dogmatically into families (Gattungen) rather than
acknowledging that these epochs can be classified variously according to dif­
ferent family resemblances. In this capacity, the notion is part of Wittgen­
stein's general resistance to dogmatism (BT 259-60; EPB 158), and linked to
the idea that an OVERVIEW constructs connecting links between the phenom­
ena it describes. It also occurs briefly in his discussion of ASPECT-PERCEPTION:
to recognize a family resemblance between different faces is the dawning of
an aspect (PI II 193, 210; RPP II §§551-6; LW I §692).
The notion is crucial to Wittgenstein's attack on essentialism, the view
that there must be something common to all instances of a concept that
explains why they fall under it (PG 74-5), and that the only adequate or
legitimate EXPLANATION of a word is an analytic definition which lays down
necessary and sufficient conditions for its application, entailing that, for
example, explanations by reference to examples are inadequate. Wittgen­
stein condemns this 'contemptuous attitude towards the particular case' as
based on a misguided 'craving for generality' (BB 17-18). The Tractatus had
succumbed to this craving in trying to delineate the essence of symbolic
representation, and in particular in its doctrine of the GENERAL PROPOSITIONAL
FORM, according to which all propositions depict possible states of affairs and
are of the form 'Things are thus-and-so.' By contrast, Philosophical
Investigations §§1-64 elucidates the concepts of language and of a proposition
with the help of a series of language-games. Wittgenstein's interlocutor com­
plains that although Wittgenstein has nattered on about language-games, he
has not stated what a language-game is, and has therefore failed to explain
the essence of language (PI §65). Wittgenstein pleads guilty to the charge,
but rejects the underlying demand on the grounds that there is no essence
of language, but only different phenomena related in various ways.
He illustrates this first by reference to the notion of a game, because of
the antecedent comparison of language to a game (PI §§66-7). When we
120
FAMILY RESEMBLANCE
'look and see" whether all games have something in common, we notice that
they are united not by a single common defining feature, but by a complex
network of overlapping and criss-crossing similarities, just as the different
members of a family resemble each other in different respects (build, fea­
tures, colour of eyes, etc.). What holds the concept together and gives it its
unity is not a 'single thread' running through all cases, but, as it were, an
overlapping of different fibres, as in a rope (BB 87; PG 75). This can be
illustrated as follows:
GAMES
A B C D E F G
F
1
E
A
2
T
U 3
R
E 4
S
5
Wittgenstein does not maintain that games have nothing in common - he
refers to them as 'procedures', and it is manifest that they are all activities.
But this falls short of a definition, since there are many activities which are
not games. The claim is that there is no set of conditions which all and only
games satisfy, and hence no analytic definition of 'game' in terms of neces­
sary and sufficient conditions. Wittgenstein presents this finding as 'the
result' of an examination (PI §66; TS302 14). But he has only argued for it
by counter-examples to some plausible definitions. He is therefore open to
the charge that, with persistence, 'game' can be analytically defined, for
example, as a rule-guided activity with fixed objectives that are of little or
no importance to the participants outside the context of the game. It could
be claimed that such a definition does not merely sharpen our concept
through a stipulation, a possibility Wittgenstein concedes (PI §69), but cap­
tures how we already use 'game' (it also captures the German Spiel, but not
spielen (playing), which covers activities without fixed rules or objectives, such
as throwing a ball in the air).
These qualms about the claim that games have no common defining
characteristics leave intact the more modest claim that they need not have
any such thing in common (BB 25, 86-7; BT 16-20, 86-7; PG 74-6; PLP
180-90). This suffices to resist the essentialist position that there must be an
analytic definition. Even if such a definition could be provided, it would not
be constitutive of the meaning of our word 'game', since the latter can be,
and has been, explained by reference to examples, not to such a common
121
 FAMILY RESEMBLANCE
characteristic. But this fall-back position also faces objections. One is that
our concept of a game is explained by reference to a common property, it is
just that this property is the disjunction of all the resemblances which link
members of the family of games. But this objection is a mere 'playing with
words' (PI §67). Unlike the suggested analytic definition, it does not provide
a standard for the correct use of 'game' other than the overlapping simila­
rities stressed by Wittgenstein. Moreover, it does not distinguish the case of
'game', in which the resemblances are themselves recognizably related, and
which therefore can be applied to an open class of new cases, from artifi­
cially constructed disjunctive concepts (e.g., of being either a member of
parliament or a cane toad).
The more serious objection is- that the notion of a family-resemblance
concept is incoherent. The proper conclusion to draw from the fact that we
explain 'game' in a variety of different ways, is that it is not a univocal
term, but has different, albeit related, meanings. Wittgenstein seems to have
rejected this suggestion, by insisting that, for example, in the case of 'under­
standing' we do not have a family of meanings, but family resemblances
within a single concept (e.g., PI §§531-2). Against him one might invoke his
own idea that the meaning of a word is its use, and that diversity of USE
entails diversity of meaning. We apply 'game' to different pairs of instances
on diverse grounds. Indeed, Wittgenstein himself intimates that a term is
ambiguous if and only if in one and the same context it can make for both
a true and a false statement (BB 58). Yet, on the account just given, saying
for example that the Olympic Games are games can be true or false
depending on the rationale applied. Wittgenstein replies to such qualms by
noting that there need not be any justification for including something
under the concept: 'a transition can be made from anything to anything'
(PG 75-6). But while it is correct that no particular concept-formation is
forced on us, we distinguish between ambiguous and univocal terms, and
between a new empirical application of a term and an extension of a con­
cept, and we do this precisely on the grounds of whether or not the new
application is licensed by the original explanation.
Wittgenstein could accept this and still insist that 'game' differs from a
genuinely ambiguous terms like 'light' or 'bank' which lack the overlapping
similarities that allow one to speak of the concept of a game or number (PI
§§67-71; PG 75). One might insist that we must distinguish three different
cases - univocality, family of meanings, as with 'game', and ambiguity -
since to reduce the second case to the first stretches the notion of univocity
beyond breaking-point. However, Wittgenstein could reply that the question
of what constitutes identity or difference in meaning or concepts cannot be
answered by criteria which are as hard and fast or context-independent as
the maxim 'same concept, same marks' suggests (PI §§547-70).
Wittgenstein himself occasionally suggests that family-resemblance con-
122
FAMILY RESEMBLANCE
cepts evolve around one or more 'centres of variation', paradigmatic cases
such as football in the case of 'game', to which we relate other cases on dif­
ferent grounds (EPB 190). This would bring his conception close to Gas-
king's idea of a 'cluster-concept', although he would reject the suggestion
that we apply 'game' to non-central instances on the basis of complex calcu­
lations concerning their differentially weighted resemblances to these core
cases.
Even if an analytic definition can be provided for 'game', there remain
other candidates, such as 'art' or 'romanticism', which seem impervious to
such attempts. This is part of the reason why Wittgenstein's notion has had
such a tremendous impact on discussions of the question 'What is art?' in
AESTHETICS. Similar considerations would apply to labels such as 'science',
'polities', 'law'. Wittgenstein also treated as expressing family-resemblance
concepts terms which are even more specific than 'game', such as 'reading',
'comparing' and even 'chair' (PI §164; BB 86-7; PG 118).
It has been suggested that the notion of family resemblance is supposed to
provide a general solution to the problem of universals: against nominalism,
it points out that different instances have more in common than merely
being called 'F; against realism, that what they share is just their being F,
not an additional common property. But this proposal distorts the idea of
common property: being F is not a property by virtue of which something
qualifies as being F (although it may be a property by virtue of which some­
thing qualifies as being G, etc.).
In any event, Wittgenstein did not propound the view that all concepts
are family-resemblance concepts. His account suggests rather that at least
some of the branches of a family-resemblance concept are united by neces­
sary and sufficient conditions. This is obvious in the case of Wittgenstein's
other paradigm of a family-resemblance concept, namely that of a NUMBER.
The various types of numbers - natural, rational, real, complex, etc. -
cannot be defined by a common property. Indeed it would even be mistake
to suppose that the natural numbers are simply a subset of the signed inte­
gers, since positive rational integers are subject to different rules - we can
subtract 9 from 5 if we are operating within the signed integers, but not if
we are operating within the natural numbers. We are dealing with a family
tree which can be variously extended. But each such extension is precisely
defined (PG 70; PI §135; for a similar idea see Russell, Introduction 63-4).
Equally, there are analytic definitions for some scientific (PLP 93-4, 183)
and legal concepts.
Family-resemblance concepts are not the only ones which do not fit the
essentialist model. Others are colour concepts, and concepts like 'high' and
'deep': 'blue' refers to a range of shades, but there is no feature which all of
these have in common by virtue of which they are blue (BB 130-7; PI
§§380-1). But Wittgenstein's main concern is with two other types of con-
123
 FIRST/THIRD-PERSON ASYMMETRY
cepts. Some passages suggest that psychological concepts, notably that of
UNDERSTANDING, are family-resemblance concepts (e.g. BB 19-20, 32~3, 115-
25, 144-52; PI §§236, 531-2; Z §26). However, this view recedes into the
background. Perhaps Wittgenstein realized that what is united by over­
lapping similarities here are the forms of behaviour on the basis of which we
ascribe such terms to others, and that this does not entail that the terms
themselves are family-resemblance terms.
The other group are the formal or categorial concepts of the Tractatus, in
particular 'proposition' and 'language' (PI §§65-8, 108, 135, 179; BT 60-74;
PG 112-27). Wittgenstein claims that no analytic definition will fit these
terms. For these are not technical terms, but terms of ordinary language,
which in their ordinary use refer to a variety of different but related phe­
nomena. Any analytic definition of such terms would be stipulative, and
would not remove the philosophical puzzles, which arise out of our ordin­
ary, unsanitized concepts (BB 25-8; PG 119-20).
Some readers have felt that abandoning the quest for analytic definitions
or for accounts which subsume phenomena under general principles is at
odds with the very idea of rational investigation. But as Aristotle has taught
us, one should not treat any topic with greater systematicity than it allows.
In so far as Wittgenstein's methodological maxim 'I'll teach you differences!'
is based on Buder's motto 'Everything is what it is, and not another thing',
it is unassailable (RW 157; PR 196; LC 27). However, it is just as dogmatic
to deny uniformity where it exists as to insist on it where it is absent.
first/third-person asymmetry see AVOWAL; INNER/OUTER; PRIVACY
form of life (Lebensform) A work by Spranger bears thetideLebensformen,
but this refers to types of individual character. Wittgenstein's term, by con­
trast, stresses the intertwining of culture, world-view and language. He may
have picked up this idea from Spengler (Decline of the West I 55), but it has a
long tradition in German philosophy (Hamann, Herder, Hegel, von Hum­
boldt). Although the term occurs only half a dozen times in Wittgenstein's
published work, it has given rise to a multitude of misinterpretations, partly
due to his nonchalant use. The term 'language-game' is meant to highlight
that 'the speaking of language is part of an activity, or of a form of life' (PI
§23; see RFM 335; MSI 19 148). Like speech-act theory, Wittgenstein stresses
that speaking is a rule-guided activity. But he goes further by holding that
our LANGUAGE-GAMES are 'interwoven' with non-linguistic activities, and must
be understood within this CONTEXT. This holds not just for our actual
speech-patterns. Indeed, the best argument for Wittgenstein's claim that the
non-linguistic context is essential to understanding linguistic activities is that
fictitious language-games can only be properly assessed if one tells a story
about how they fit in with the overall practice of the fictitious community.
124
FORM OF LIFE
'To imagine a language means to imagine a form of life' (PI §§)^ 19). In
Blue and Brown Books 134, to imagine a language is equated with imagining a
'culture'. Accordingly, a form of life is a culture or social formation, the
totality of communal activities into which language-games are embedded.
At the same time, Wittgenstein also speaks of forms of life. 'Instead of the
unanalysable, specific, indefinable: the fact that we act in such-and-such
ways, e.g. punish certain actions, establish the state of affairs thus-and-so, give
orders, report, describe colours, take an interest in the feelings of others.
What has to be accepted, the given - one might say - are facts of life//
forms of life' (RPP I §630; MS 133 54). This passage has been invoked to
show that a form of life is a language-game, and that there are coundess
forms of life, just as there are language-games. But even leaving aside the
singular use noted above, the facts of life listed are not uniformly language-
games. Rather, facts of life are the specific patterns of behaviour which
together constitute a form of life.
'It is characteristic of our language that the foundation on which it
grows consists^ in steady forms of life, regular activity. Its function is deter­
mined above all by the action which it accompanies' (CE 404). These
remarks shed badly needed light on the famous What has to be accepted,
1
the given, is — so one could say - forms of life (PI II 226). In the Tractatus,
the foundations of language were provided by 'unanalysable' sempiternal
OBJECTS, whose essences — combinatorial possibilities — are supposed to
determine, in an ineffable way, the LOGICAL SPACE of possible situations, and
thereby set unalterable limits to what it makes sense to say. Now Wittgen­
stein holds that in so far as language has foundations, they are provided
not by metaphysical atoms (see PR 72), but by shifting patterns of commu­
nal' activity.
The idea that forms of life provide the foundations of language has been
elaborated in two opposite directions. On a transcendental reading, the notions
of a language-game and of a form of life take the place of the (quasi-)trans-
cendental preconditions of symbolic representation in the Tractatus. How­
ever, even if our communal practice is a precondition of our language-
games, this does not amount to a justification (transcendental deduction) of
that practice (although the fact that language requires the context of a prac­
tice may reveal certain sceptical doubts to be nonsensical). Furthermore,
although the conditions of sense laid down by GRAMMAR antecede matters of
fact decided by reality, the point of the notion of a form of life is precisely
to de-transcendentalize that contrast by acknowledging that grammar is an
integral part of human practice, and hence subject to change.
The opposing interpretation is naturalistic. It is often held that our form of
life is part of our inflexible biological human nature which rigidly deter­
mines how we act and react. This might be supported by reference to Witt­
genstein's claim that he provides 'remarks on the natural history of human
125
 FORM OF LIFE
beings' (PI §415). However, Wittgenstein's naturalism is anthropological
rather than biological. Ordering, questioning, recounting, chatting are 'as
much a part of our natural history as walking, eating, drinking, playing' (PI
§25). These activities, as well as those already quoted, are cultural activities,
forms of social interaction. Again, measuring, and even mathematics and
logic, are 'anthropological phenomena' which are part of our 'natural his­
tory' (RFM 352-3, 356, 399; RPP I §1109). This natural history is the his­
tory of cultural, language-using creatures. We must distinguish forms of life
from the common human nature onto which they are grafted (see FRAME­
WORK). Wittgenstein (like Marxism and pragmatism) stresses not our inflex­
ible biological outfit, but our historical practice.
Equally, it has been suggested that there is really only one form of life for
human beings, that different forms of life are simply unintelligible to us: it is
indeed a contingent fact that we speak and act as we do; and we may even
be in a position to appreciate that it is logically possible to have different
forms of life; but human nature prevents us from understanding these alter­
native forms of life themselves. This is at odds with Wittgenstein's insistence
that different FORMS OF REPRESENTATION become intelligible against the back­
ground of different forms of life. Measuring with elastic rulers (RFM 38, 91-4)
is no different from measuring by the ell. It makes good sense for a commu­
nity with concerns different from ours. To be sure, there is a difference
between recognizing that people measured by the ell in the Middle Ages
and imagining that we could revert to this technique now. Such a change
would involve fundamental changes in our techniques, our technology and
hence our goals and values. But it is not unintelligible; we can understand
what is involved, even though it does not appeal to us. Different forms of
representation are intelligible given different training or different purposes (Z
§§352, 387-8). Even the idea that they must fulfil something we recognize as
a relevant purpose is merely a prejudice of our instrumental form of life (see
RPP I §49; RFM 95).
As regards linguistic practices, Wittgenstein embraces not a naturalist
determinism, but a cultural relativism (e.g. MS 109 58), which follows from
the conceptual relativism of the AUTONOMY OF LANGUAGE. The latter denies
merely that our forms of representation are subject to metaphysical stan­
dards, a putative essence of reality, not that they may be subject to prag­
matic standards. However, if is based on the idea that each form of
representation lays down its own standards of rationality, which implies that
even pragmatic justifications are internal to particular language-games.
Hence, criticizing a 'language-game' from the outside can never be a matter
of rational argument, but only of 'persuasion' (OC §§92, 262, 608-12; see
CERTAINTY). Note, however, that language-game relativism should not be
Wittgenstein's final word. Within the framework of a form of life it is possi­
ble to justify or reform particular language-games - a grammatical proposi-
126
FORM OF LIFE
tion like 'One cannot know the future' might be justified by the unreliability
of our predictions (LW I §188), or reformed if their reliability improved
drastically. What cannot be so criticized is the linguistic practice (form of
life) as a whole.
Like other relativists, Wittgenstein studiously ignores the objection that his
position is self-refuting because it is implicidy committed to the claim that it
itself is correct in a way it explicidy rejects. Unlike other relativists, Wittgen­
stein might have a reply. His points about the immanence of justification and
doubt do not apply epistemic terms in a way limited by the practice con­
cerned. They are grammatical remarks, reminders of the way these words are
used in this practice. As such they aspire to be correct in a way which trans­
cends different practices - they could be made by a philosopher who engages
in a different practice. But this is compatible with acknowledging that there is
no necessity about engaging in a particular language-game. Wittgenstein can
be a conceptual relativist, but not a philosophical one.
Wittgenstein can be accused of ignoring the fact that in justifying, for
example, our scientific world-picture against a community that predicts the
future on the basis of oracles (OC §609), we can draw on certain universal
values, such as due respect for experience and successful prediction. If it
turns out that meteorology is better at predicting the weather, a community
which persistendy sticks to oracles can be accused of instrumental irration­
ality. However, this does not mean that they must abandon their practice,
since their adherence may express different priorities. Another possibility
which Wittgenstein himself mentions is that we may be able to claim that
our world-view encompasses theirs and is hence richer (OC §286). But in
other respects our scientific and technological world-view may actually be
impoverished.
Perhaps Wittgenstein never came to explore the rational limitations to
relativism because he increasingly stressed the naturalistic limitations. He
regards his certainty that there is, say, a chair over there 'not as something
akin to hastiness or superficiality, but as a form of life . . . as something that
lies beyond being justified or unjustified; as it were, as something animal'
(OC §§358-9). But he adds '(That is very badly expressed and probably
badly thought as well).' What is inflexible are not forms of life in the sense
of social practices, but some of their constituent activities or facts of nature.
We could not stop taking an interest in people's pain (LW II 43). 'Language
... is a refinement, in the beginning was the deed. First there must be a
firm, hard stone for building . . . Afterwards it's certainly important that the
stone can be trimmed, that it's not too hard' (CE 420; CV 31). The point is
not so much that human nature is immutable as that language in general
and reasoning in particular are rooted in forms of behaviour which are nei­
ther rational nor irrational, but antecede questions of rationality (OC §§204,
475).
127
 FORM OF LIFE
Anticipating the current debate about radical translation, Wittgenstein
briefly discussed the 'ethnological point of view' or 'anthropological method'
which we adopt when coming to understand an (actual or invented) alien
community (CV 37; SDE 25). Like Quine and Davidson, he insists that
there are rninimum requirements which a form of linguistic behaviour must
meet in order to be intelligible to us. According to their 'principle of charity',
interpretation presupposes that we can treat the aliens' beliefs as by and
large true. Wittgenstein concurs partly. 'If language is to be a means of com­
munication there must be agreement not only in definitions but also . . . in
judgements' (PI §242). But while the principle of charity stresses the second
point, it discards the first. By maximizing agreement in opinion, it puts the
cart (truth) before the horse (meaning). By and large, we must understand
what people say in order to judge whether they are speaking the truth. Shar­
ing a language is 'not agreement in opinions but in form of life' (PI §241; see
RFM 353). By the same token, understanding an alien language presupposes
convergence not of beliefs, but of patterns of behaviour, which presuppose
common perceptual capacities, needs and emotions: "The common beha­
viour of mankind is the system of reference by means of which we interpret
an unknown language' (PI §206; see RFM 414-21; EPB 149).
This explains Wittgenstein's puzzling remark that 'If a lion could talk, we
could not understand him' (PI LI 223). On one reading, this means that we
could not understand a lion who utters English sentences like 'I'm not inter­
ested in you, I've just had an antelope', which is obviously false (although
one might, following Austin, question whether such a talkative creature
could count as a lion). On a charitable reading, it means that if lions had a
feline language of complex growls, roars, etc., we could never come to learn
it. Why? Because their form of life, and their behavioural repertoire, are so
alien to us. We could not make head or tail of their facial expressions, ges­
tures and demeanour. Moreover, our ability to interact even with a tame
lion is stricdy limited. For related reasons we 'could not find our feet' with a
community of human beings who give no expression of feeling of any kind,
and we would be completely at a loss with spherical Martians (Z §390; LC
2-3; see also RPP H §568; LW I §190; MS137 13.1148).
The need for convergence in form of life has yet unexplored implications
for ethics. It might be used to justify the idea that our obligations towards
living human beings are of a different kind from those towards animals,
simply because our ability to interact with animals, to share ideas, responsi­
bilities and aspirations with them, is so severely restricted.
Wittgenstein's form of life contextualism became more pronounced with
time. He claimed that to describe human action we need to describe not
just what 'one man is doing now, but the whole hurly-burly of human
actions', the 'way of living' of which an individual action is part (Z §567;
RFM 335-6). Sensation-terms like 'pain' are applied to others on the basis
128
FORM OF REPRESENTATION
of straightforward behavioural CRITERIA. By contrast, moods and intentional
attitudes (hope, pretence, grief, INTENDING, RULE-FOLLOWING) cannot be ascri­
bed simply on the basis of an individual's momentary behaviour, but require
a certain surrounding. This 'context' is provided not by certain mental
accompaniments, but by (a) the subject's abilities; (b) the 'whole history of
the incident', by what went on before and after; (c) the social surroundings,
that is, the existence of certain language-games in the subject's linguistic
community. For example, a baby moving a chess-piece does not count as
playing chess, nor is a baby capable of pretending. One can be in pain for
a split second, but not for so brief an instant expect someone, or be in grief.
And one can intend to play chess only if the technique of the game exists
(PI §§200, 205, 250, 337, 583, 643-4; BB 147; RPP II §631; LW I §§859-
76; LW LI 26-47; Z §99).
After Philosophical Investigations Part I, Wittgenstein expressed this by saying
that such terms refer to 'patterns in the weave of our life' (Lebensmuster) (PI II
174, 229; LW I §§862-9, 942, 966; LW LL 42-3, 55-6, 84; Z §§567-9). The
complexity of this weave explains why some third-person psychological
judgements are uncertain. The possibility of disagreement about the emo­
tions of others reflects an indeterminacy which is constitutive of some of our
psychological terms. That indeterminacy in turn is due to communal pat­
terns of behaviour: mental concepts must be elastic and flexible because
human behaviour, and our reaction to it, is diverse and unpredictable (RPP
II §§651-3; LW I §§206-11; LW LI 24-5, 6 H , 72, 84-95). We cannot
make subde emotional ascriptions on the basis of simple criteria, but need to
take into account the context and previous events. This is often possible
only if one is acquainted with the person concerned, and has intimate
knowledge of human nature.
form of representation (Form der Darstelhing) In the Tractatus, this is the
external 'standpoint' from which a picture represents its subject (TLP
2.173L; see LOGICAL FORM). Closer to Wittgenstein's later notion is the Hert­
zian idea that different scientific theories are guided by different 'forms of
describing the world' (Formen der Weltbeschreibung). The later Wittgenstein
extends this idea beyond SCIENCE. A 'representational form' is a way of look­
ing at things, a kind of Weltanschauung (PI §122). This idea encompasses
one's approach to philosophy, which in Wittgenstein's case is guided by the
attempt to provide an OVERVIEW of grammar. By contrast, the world-picture
(Weltbild) of On Certainty is the inherited background of our scientific and
everyday reasoning. Like a 'mythology', it can itself be altered not through
reasoning, but only through a conversion (OC §§92, 94-7, 167, 262, 612).
Similarly, Wittgenstein characterizes GRAMMAR, the system of rules which
provide standards for the correct use of words, as our 'method' or 'form of
representation' (M 51; OC §§61-2; PI §§50, 104, 158). 'That one proposi-
129
 FORM O F REPRESENTATION
tion is true and another false is no part of grammar. What belongs to gram­
mar are all the conditions (the method) necessary for comparing the propo­
sition with reality. That is, all the conditions necessary for the understanding
(of the sense)' (PG 88). In virtue of determining what combinations of signs
make sense, and hence count as candidates for truth — 'the kind of statement
we make about phenomena' (PI §90) - grammar itself is not subject to
empirical refutation. Logic is 'antecedent' to the correspondence between
'what is said and reality' (RFM 96).
This provides the key to Wittgenstein's later account of logical necessity.
As in the Tractatus, he resists the Platonist view of necessary propositions as
part of a super-physics of the abstract which differ from empirical proposi­
tions merely by describing more abstract objects. He also rejects the empiri­
cist reduction of necessary propositions to empirical generalizations (see
INTERNAL RELATIONS). The contrast between them is even greater than tradi­
tionally assumed. Empirical propositions can be said to describe possible
states of affairs, but necessary propositions cannot be said to describe neces­
sary states of affairs. For their role is normative rather than descriptive.
They function as, or are linked to, 'grammatical propositions', sentences
which are typically used to express grammatical rules. A grammatical propo­
sition like
(1) Black is darker than white
is a 'norm of description' or of 'representation' (RFM 75-6; AWL 16; OC
§§167,. 321). It lays down what counts as an intelligible description of reality,
establishes internal relations between concepts ('black' and 'white') and licen­
ses transformations of empirical propositions (from 'Coal is black and snow
is white' to 'Coal is darker than snow').
Grammatical propositions antecede experience in an innocuous sense (PR
143; LWL 12; AWL 90). They can be neither confirmed nor confuted by
experience. (1) cannot be overthrown by the putative statement 'This white
object is darker than that black object', since the latter is a nonsensical com­
bination of signs. This antecedence to experience renders intelligible the
apparendy mysterious 'hardness' of necessary propositions and internal rela­
tions (PI §437; RFM 84; PG 126-7). To say that it is logically impossible for
a white object to be darker than a black one is to say that we would not call
an object both 'white' and 'darker than a black object'. Given our rules, it
makes no sense to apply both terms to one and the same object at the same
time. Wittgenstein explains logical necessity through the distinction between
sense and nonsense drawn by our norms of representation.
As in the Tractatus, Wittgenstein emphasizes the differences between var­
ious kinds of necessary propositions. He holds on to his earlier account of
logical propositions as TAUTOLOGIES (AWL 137-40; LFM 277-81). But he no
longer condemns other necessary truths as pseudo-propositions. Arithmetical
130
FORM OF REPRESENTATION
equations, geometrical propositions and analytic propositions are gramma­
tical rules (see respectively WVC 156; PG 347; RFM 363 and WVC 38,
61-3; LWL 8, 55 and PI §251). Metaphysical propositions, if they are not
merely nonsensical, typically mask grammatical rules (BB 35, 55; AWL 65-9;
Z §458). Their linguistic appearance is that of statements of fact, but their
actual role is that of grammatical propositions.
Unlike their predecessors (rules of LOGICAL SYNTAX), grammatical rules are
'conventions' (Ubereinkunft, Konvention). Although they are rarely subject to
decisions their function, if not their history, is that of conventions (PI §§354-
5; AWL 89-90, 156-7; PG 68, 190). Grammar is AUTONOMOUS, it reflects
neither the essence of reality nor an inflexible human nature (see FRAME­
WORK). Accordingly, Wittgenstein's account of logical necessity is con­
ventionalist. However, it differs substantially from the conventionalism of the
logical positivists. Their goal was to develop a form of empiricism that could
account for logical necessity without either reducing it to empirical general­
ity, lapsing into Platonism or admitting synthetic a priori truths. Necessary
propositions, the positivists argued, are a priori, but do not amount to
knowledge about the world. For, with the help of the Tractatus, it seemed
that all necessary propositions could be seen as dhalytic, true solely in virtue
of the meanings of their constituent words. Logical truths are tautologies
which are true in virtue of the meaning of the LOGICAL CONSTANTS alone, and
analytical truths can be reduced to tautologies by substituting synonyms for
synonyms - thus 'All bachelors are unmarried' is transformed into 'All
unmarried men are unmarried', a tautology of the form \x)((Jx. gx) Z) gx)'
the truth of which follows from the meaning of the logical signs involved.
Necessary propositions are true by meaning or convention. They either are
themselves conventions (definitions) or follow from such conventions.
Wittgenstein's distinction between grammatical and empirical propositions
deviates from the logical positivists' analytic/synthetic distinction in four
respects, (a) Many of his grammatical propositions do not fit into even the
most generous list of analytical truths. The reason is that Wittgenstein had
realized that there are non-truth-functional logical relations (PR 105-6), and
hence necessary propositions like (1) which are not analytic in the sense of
the Tractatus and the Vienna Circle, (b) The analytic/synthetic distinction is
set up in terms of the forms and constituents of type-sentences. But whether
an utterance expresses a grammatical proposition, that is, is used to express
a linguistic rule, depends on its role on an occasion of utterance, on whe­
ther in the particular case it is used as a standard of correctness. For exam­
ple, 'War is war' is not typically used to express the law of identity (PI II
221; WVC 153-4; PR 59; AWL 64-5; BT 241). (c) The distinction involves
the idea that the truth of necessary propositions is a consequence of the mean­
ing of their constituents. According to Wittgenstein, necessary propositions
determine rather than follow from the meaning of words, since they are partly
131
 FORM OF REPRESENTATION
constitutive of the meaning of the constituent terms (see MEANING-BODY), (d) By
explaining the status of necessary propositions by reference to their norma­
tive rather than descriptive employment, Wittgenstein rejects the view that
they are a special kind of truths, one whose source is meaning or conven­
tion instead of experience. Notably, if tautologies are degenerate proposi­
tions which do not say anything, a point the logical positivists accepted, in
what sense could they be true?
These differences are brought out by the fact that Wittgenstein con­
templates using the label 'synthetic a priori' for, firsdy, mathematical propo­
sitions, presumably because they can be used descriptively as well as
2
normatively ('25 = 625' can be used as a prognosis of what result people
will get when they square 25, though it is in fact used as a criterion for
having performed that operation - RFM 318-19, 327-30); and secondly,
grammatical propositions which cannot be explained through the predicate
calculus - 'There is no reddish-green' or '"Above" has five letters', for
example (RFM 245-6, 336). Kant's idea that mathematical and metaphysi­
cal propositions are synthetic a priori expresses an insight: the fact that they
seem to anticipate reality requires an explanation. The Viennese account of
all necessary propositions as truths which say nothing leaves them without a
role. In repudiating the separation of necessary propositions from their
application, Wittgenstein takes up Kant's problem. But he insists that neces­
sary propositions are a priori precisely because they are not about anything
and hence not synthetic (WVC 67, 77-8; LWL 79; PLP 67-8). The role of
necessary propositions for empirical discourse is that of norms of representa­
tion which provide guidelines for 'charmelling' (dealing with) experience
(RFM 240, 324-5, 387). 'Whenever we say that something must be the
case, we are using a norm of expression'; a logical connection 'is always a
connexion in grammar' (AWL 16, 162; RFM 64, 88).
While Wittgenstein's conventionalism avoids the difficulties of the Vien­
nese version, it faces serious ones of its own (see MATHEMATICAL PROOF). Even
sympathetic commentators like Waismann have felt that the claim that
necessary propositions are rules ignores that the former are about numbers,
colours, lengths, sensations, etc., not words; and that the former but not the
latter can be said to be true (PLP 66-7, 136-7). However, Wittgenstein could
grant that necessary propositions are not in fact rules while insisting that
they resemble rules in that they 'play the role of norms of description'
(RFM 363; LFM 55, 256) - they license transformations of empirical propo­
sitions. Moreover, his point is that for a necessary proposition to be about
something and to be true is toto caelo different from what it is for an empiri­
cal proposition to be so (AWL 154; LFM 114, 250-1; PI §251). The role of
a grammatical proposition like 'All bachelors are unmarried' is not to make
a true statement of fact about bachelors but to explain the meaning of
'bachelor'. We do not verify it by investigating the marital status of people
132
FORM OF REPRESENTATION
identified as bachelors, and its denial displays not factual ignorance but
linguistic misunderstanding. Most importandy, it excludes not a genuine pos­
sibility, but only a NONSENSICAL form of words.
Even if Wittgenstein's conventionalism is not wholly satisfactory, his dis­
tinction between grammatical and empirical propositions not only escapes
Quine's celebrated attack on the analytic/synthetic distinction, but also
helps to undermine Quine's empiricist assimilation of necessary and empiri­
cal propositions. Because of (c), it avoids what Quine calls the 'myth of the
museum', the idea that abstract entities (logical forms or meanings) force us
to hold on to certain propositions come what may; and because of (d) it is
not committed to the idea of 'truth by convention'. Moreover, Wittgenstein
can accommodate Quine's holistic picture of a web of beliefs according to
which even 'necessary propositions' can be abandoned to preserve other
beliefs. Indeed, he himself advocated such a holism during his VERIFICA-
TIONIST period: 'hypotheses', that is, all propositions going beyond what is
immediately given to the senses, cannot be conclusively verified or falsified,
because recalcitrant evidence can be accommodated by auxiliary hypotheses
(PR 285-90). This may have influenced Catnap's holism in The Logical
Syntax of Language, and hence, indirecdy, Quine's*own. During the transition
period, Wittgenstein did not extend this reusability to necessary proposi­
tions, and he later dropped the empiricist myth of unconceptualized sense-
experiences. But the holistic picture survives in On Certainty (OC §§94-6,
512-19). Moreover, his functional conception of grammatical rules, accord­
ing to which an expression is a rule if it is employed as a standard of cor­
rect use, implies that the logical status of sentences can change according to
our way of using them. Empirical propositions are 'hardened' into rules
(RFM 325, cp. 192, 338-9), while rules lose their privileged status and are
abandoned. For example, the sentence 'An acid is a substance which, in
solution, turns litmus-paper red' lost its normative status (acids now being
defined as proton-donors) and turned into an empirical statement which
holds true of most, but not all, acids. Conversely, the statement 'Gold has
79 protons' was originally an empirical discovery but is now partly con­
stitutive of what we mean by 'gold'.
Unlike Quine, but like Carnap, Grice and Strawson, Wittgenstein insists
that this is compatible with a dynamic distinction between necessary and
empirical propositions. The abandoning of grammatical propositions can be
motivated by theoretical considerations ranging from new experiences to
simplicity, fruitfulness or sheer beauty. But it is distinct from the falsification
of a theory. There is no such thing as the falsification of a grammatical pro­
position. For its normative status means that the proposition itself is (partly)
constitutive of the meaning of its constituent terms (BB 23, 56; AWL 40).
After such a revision, it makes sense to use words in ways which were pre­
viously excluded as nonsensical.
133
 FORM OF REPRESENTATION
(2) Nobody under the age of 10 is an adult
is a grammatical proposition which partly determines whom we call an
adult. If we were to allow a statement like
(3) Jane's three-year-old daughter is an adult
because, for example, she has amazing intellectual capacities, we would not
have falsified (2). For allowing (3) amounts to instituting a new way of using
'adult', and this introduces a new concept. Consequentiy (2) and (3) would
not contradict each other, since 'adult' means something different in each
case.
Scientific concepts are typically held in place by more than one explana­
tion. In cases where several phenomena (fever, presence of virus) are found
together in association with a particular disease, the only way to distinguish
between CRITERIA and symptoms may be a decision (BB 25). 'The fluctuation
in grammar between criteria and symptoms makes it look as if there were
nothing at all but symptoms' (PI §354, cp. §79; Z §438). But with respect to
specific experiments, it is often possible to decide whether or not particular
statements are used normatively or empirically. To deny this would be to
deny that one can distinguish, with respect to a particular measurement,
between the role of the ruler and the role of the object measured (PI §50).
Indeed, a collection of beliefs can only be woven into a web if certain pro­
positions are not merely abandoned with greater reluctance than others, but
play a different role, namely that of establishing logical connections between
different beliefs (this is of a piece with Lewis Carroll's insight into the need
to distinguish between the axioms and the rules of inference of a formal
system).
Wittgenstein anticipated Quine's assimilation of necessary propositions to
well-entrenched beliefs (presumably because he saw it as the inevitable con­
sequence of Russell's and Ramsey's empiricist conception of mathematics)
but claimed that it ignores 'the deep need for the convention' (RFM 65, 237).
[TJf there were only an external connection no connection could be de­
scribed at all, since we only describe the external connection by means of
an internal one. If this is lacking we lose the footing we need for describ­
ing anything at all - just as we can't shift anything with our hands unless
our feet are planted firmly. (PR 66)
If all the norms of representation concerning, for example, 'bachelor' were
transformed into empirical propositions this would mean that all of the fol­
lowing sentences could be rejected: 'Bachelors are unmarried men', 'Bache­
lors are human beings', 'Bachelors are made of flesh and blood.' Under
these circumstances anything at all could be called 'bachelor', since there
134
FRAMEWORK
would be no reason to deny that anything falls under the concept. Conse­
quentiy, the use of this term would have become totally arbitrary, that is,
the term itself would have become senseless. Correspondingly, if we surren­
dered the grammatical rules governing the use of all our words, these words
would lose all meaning. Of course, our habit of uttering words might con­
tinue: a communal phonetic babbling without rules is conceivable. But it
would resemble speaking in tongues more than a 'language' (PI §§207, 528).
If anything can be said, nothing can be meaningfully said.
There is an important parallel between Wittgenstein and Quine. Both
characterize logical truths not in terms of their form or structure, but by
reference to linguistic behaviour. But unlike Quine's reductionist behaviour­
ism, Wittgenstein views language as essentially guided by norms. It is this
normativist conception of language which allows him to make sense of,
rather than to reject, the notion of logical necessity.
formal concepts see SAYING/SHOWING
framework One of the principles of Wittgenstein's early philosophy was
the autonomy of sense: whether a proposition make? sense must not depend
on another proposition's truth (NM 117; TLP 2.0211). Language is a self-
contained abstract system governed 6y rules of LOGICAL SYNTAX. Recognizing
the importance of the surroundings of language is a major achievement of
Wittgenstein's later reflections. His first step is to radicalize the Tractatus's
CONTEXTUAUSM: a word has meaning only as part of a LANGUAGE-GAME, which
itself is part of a communal FORM OF LIFE. The second is a kind of
naturalism. Our linguistic and non-linguistic activities are conditioned by
certain 'facts of nature'. Our concepts rest on a 'scaffolding of facts' in that
different facts of nature would make inteHigible different 'concept-forma­
tions' (PI II 230; RPP I §48; Z §§350, 387-8). In this context Wittgenstein
distinguishes three elements:
(a) the GRAMMATICAL rules which constitute a language-game like that of
measurement;
(b) the application of these rules in empirical propositions (specific mea­
surements);
(c) the framework or 'scaffolding' which allows us to operate the lan­
guage-game.
Disputes do not break out . . . over the question whether a rule has been
obeyed or not . . . That is part of the scaffolding from which our language
operates . . . [Human beings] agree in the language they use. That is not
agreement in opinions but in form of life. If language is to be a means of
communication there must be agreement not only in definitions but also
135
 FRAMEWORK
. . . in judgements. This seems to abolish logic, but does not do so. It is
one thing to describe methods of measurement, and another to obtain
and state results of measurement. But what we call 'measuring' is partly
determined by a certain constancy in results of measurement. (PI §§240-
2; see OC §156)
This passage can be rendered consistent if 'agreement . . . in form of life' is
not exhausted by agreement in definitions/judgements (i.e., opinions), but
includes 'a consensus of action', of applying the same technique (LFM 183—
4). The idea that language requires agreement in judgements as well as defi­
nitions would abolish logic, if communal consensus determined whether or
not a particular measurement is correct. This is why Wittgenstein insists that
what counts as the correct application of rules (an accurate measurement) is
determined by the rules themselves, which are pur standards of correctness;
the definition of 'correct measurement' is not 'what people agree on'. These
rules specify neither the results of particular measurements - ( b ) - nor that
there is general agreement in applying them - (c) (RFM 322-5, 359-66,
379-89, 406-14; Z §§319, 428-31; see RULE-FOLLOWING; TRUTH). Nevertheless,
without such agreement, the rules would 'lose their point' (PI §142; RFM
200); a technique which did not produce such consensus would not be
called 'measuring' (according to Wittgenstein, therefore, in this exceptional
case the rules themselves do include a reference to consensus).
The required consensus in application is less stringent for emotion-terms,
for example (LW II 23-4; PI II 224-8), and minimal for essentially con­
tested terms like 'corrupt'. Moreover, communal agreement is not the only
framework or background condition for playing certain language-games.
Thus, our concepts of measures work only in a world with relatively stable
rigid objects; but this is not laid down in the rules of, for example, metric
measurement. What Wittgenstein calls 'facts of nature' play the same role
(although sometimes by allowing consensus). These facts fall into three
groups:
General regularities concerning the world around us. Objects do not van­
ish or come into existence, grow or shrink, etc., in a rapid or chaotic
manner (PI §142).
Biological and anthropological facts concerning us. Our perceptual capa­
cities allow us to discern such-and-such colours (Z §§345, 368; PLP
250-4), our memory permits calculations of a certain complexity
(MSI 18 131), our shared patterns of reaction allow us to teach (AWL
102; LFM 182) - OSTENSIVE DEFLNTTION, for example, presupposes that
human beings look not at the pointing finger (as cats do), but in the
direction in which it points.
Socio-historical facts concerning particular groups or periods. Our ways
136
FRAMEWORK
of speaking express practical needs and interests (RFM 41, 80-1) shaped
by history.
Given these facts, certain forms of representation will be 'practical' or
'impractical' (AWL 70). Provided that the world is as it is, people who
employed alternative ways of calculating or measuring for purposes similar
to ours would have to make tedious adjustments. By the same token, drastic
changes in these facts could render our rules inadequate in this pragmatic
sense. They might not' only become impractical but even be inapplicable (PI
§569; RFM 51-2, 200). If objects constandy and unpredictably vanished or
sprang into existence, our language-game of counting would loose its 'point'
or become 'unusable'. So too would our colour-concepts if objects constandy
changed their colours at random.
The rules of tennis do not include that it is to be played at Earth-gravity.
But tennis would be poindess on the moon (every serve would be an ace)
and could not be played on Jupiter. Although the framework conditions do
not determine what the rules of the language-game are, they partly deter­
mine what language-games are played. Hence they impose limits on the
possibility of adopting different grammatical rules *{see AUTONOMY OF LAN­
GUAGE). 'Yes, but has nature nothing to say here? Indeed she has - but she
makes her voice audible in another way. "You'll surely run up against exis­
tence and non-existence somewhere!" But that means against facts, not con­
cepts' (Z §364). The way we speak is part of human practice, and hence
subject to the same kinds of factors that determine human behaviour in gen­
eral. However, these facts of nature do not provide a naturalistic justification
of our grammar. A change in the framework conditions would render our
rules not incorrect (false to the facts) but poindess or obsolete (PG 109-10;
Z §§366-7; RPP II §§347-53).
Wittgenstein would not even concede that given such-and-such framework
conditions we are causally forced to adopt specific language-games (Z §351).
The relative stability of the material world is a condition for measurement,
but does not force us to adopt the metric system (that is a prerogative of the
EC Commission). Similarly, common colour discriminatory abilities and the
relative constancy of the colours of things are framework conditions of any
colour grammar, but these are compatible with widely differing colour
grammars among the various languages of mankind. This is at odds with
the idea that the right, or perhaps just inevitable, rules are those which we
find natural. Wittgenstein acknowledges that we find certain rules 'natural'
(AWL 67; LFM e.g. 183, 243), but adds that this is relative to people and
circumstances; it is not biologically fixed, but malleable, for example
through education (Z §387; PI §§595-6).
Framework conditions impose causal constraints: they partly explain why
we do not go down one road, but not why we go down another. One may
137
 FRAMEWORK
feel nevertheless that acknowledging them pollutes philosophical descriptions
of grammar with causal EXPLANATIONS. Wittgenstein himself claims to supply
'remarks on the natural history of human beings' (PI §415); while elsewhere
he disavows such ambitions:
Our interest certainly includes the correspondence between concepts and
very general facts of nature. (Such facts as mosdy do not strike us because
of their generality.) But our interest does not fall back upon these possible
causes of concept-formation. We are not doing natural science; nor even
natural history - since we can also invent fictitious natural history for our
purposes. (PI H 230; see RPP I §48)
The last remark does not, however, keep philosophy free of causal hypoth­
eses, since the latter can connect fictitious background conditions with ficti­
tious concept-formations. More promising are Wittgenstein's attempts to
distinguish his kind of natural history from natural science. Sometimes they
leave unclear how it differs from straightforward grammatical remarks, for
example when he suggests that it includes such propositions as 'Grass­
hoppers cannot read or write', although not 'Human beings think, grass­
hoppers don't' (RPP II §§14-25). Equally, for Wittgenstein it is a conceptual
point that people with different discriminatory capacities could not have our
coLOUR-concepts. In other passages, however, his kind of natural history
clearly concerns 'empirical', that is, contingent facts, for example that
human beings modify their concepts in response to experience (Z §352).
Unlike grammatical reminders they do not remind us of the linguistic rules
we follow; instead, they remind us of facts about ourselves. But these empiri­
cal facts are not arcane, a topic of scientific hypotheses. The natural history
of measurement is not a branch of applied physics about how best to mea­
sure something under certain conditions. Rather, it assembles empirical facts
in a way that makes intelligible or unsurprising the one point which matters
to philosophy: that if certain contingent framework conditions changed, we
would find alternative procedures plausible or useful, and our actual proce­
dures impracticable or poindess (RPP I §§950-1109; LW I §§207-9; see
OVERVIEW). Physics might tell us that a change in certain laws of nature
would lead objects to grow or shrink constandy and chaotically. But it does
not take physics to appreciate that under such circumstances measuring sizes
would become poindess. The relevant facts go unnoticed precisely because
they are so familiar and general - a 'miss the wood for the trees' effect (PI
§129, II 230; RPP I §§46, 78).
This theme recurs in On Certainty. Wittgenstein there discusses the empiri­
cal common-sense truisms which Moore had claimed to know for CERTAIN.
He treats them as world-picture or hinge propositions: although they are
empirical, that is, state contingent facts, they could not simply turn out to be
138
FRAMEWORK
false, since this would remove the background against which we distinguish
true and false. On Certainty occasionally speaks of these propositions as a
'scaffolding' or 'framework' of our thought, and, like Phiksophkal Investigatkns,
states that 'the possiblity of a language-game is conditioned by certain facts'
(OC §§211, 617). Nevertheless, the points behind the notions of facts of
nature and hinge propositions differ in principle: if facts of nature were dif­
ferent, our language-games would change; if we could not be certain of
hinge propositions, our web of beliefs would collapse. There is an overlap
between the two categories: if certain 'unheard-of events' (OC §513) were to
occur, for example, objects grow or shrink constandy for no apparent
reason, this would not only shake our system of beliefs but, as we have seen,
render poindess or impracticable specific language-games. But uncertainty
about some hinge propositions (e.g., the spherical nature of the earth) would
affect not so much specific language-games as forms of representation within
a specialized scientific discourse.
Wittgenstein claims that hinge propositions, like facts of nature, go un­
noticed because they form part of the background of our language-games.
They are 'withdrawn from circulation' and 'shunted onto an unused siding',
and 'lie apart from the road travelled by enqutry' (OC §§88, 210). Some
commentators have concluded that hinge propositions are ghosdy phenom­
ena because they are abstract, irfeffable and transcend our linguistic prac­
tice, violating the idea that meaning is use. However, the very point of
speaking about hinge propositions is that they play a special role in our lin­
guistic practice (OC §§94-8, 152, 248). Moreover, On Certainty holds only
that hinge propositions are, by and large, not stated, not that they cannot be
stated. Wittgenstein's point is that 'if they are ever formulated', they are
exempt from doubt (OC §88). It has also been claimed that On Certainty
revives the Tractatus's SAYING/SHOWING distinction and that hinge propositions
can only show themselves in our practice. But one passage .invoked in this
context merely raises the possibility, and the other ends by stating that
'that's not how it is' (OC §§501, 618). What is correct is this. Wittgenstein
tentatively suggested that to say, with Moore, that we know hinge proposi­
tions creates confusion because it invites sceptical doubts, and is hence at
odds with our treating them as certain, which shows itself in the way we act
(e.g. OC §§7, 466). But this is not to say that it creates confusion or fuels
scepticism to draw attention to these propositions, as long as one does not
mistake them for ordinary empirical claims. Like the structure of Husserl's
'fife-world', facts of nature and hinge propositions are not ineffable, but
special: their role is too basic to be easily noted.
139

G solely upon the truth of the elementary propositions into which it can be
general propositional form (allgemeine Satzform) Wittgenstein's early
philosophy seeks to determine the nature of representation and of what is
represented, the world. It does so by establishing the essence of the proposi­
tion. Various types of propositions differ in their logical forms, and these are
to be discovered by the application of logic. But these possible forms have
something in common which is established a priori. That a form of words
can constitute a proposition is not a matter of experience, but implicit in the
rules of LOGICAL SYNTAX. The general propositional form is the essence of the
proposition, the necessary and sufficient conditions for something to be a
proposition in any 'sign-language' (Zeichensprache). Since language is the total­
ity of propositions, the general propositional form also provides features
which unite all languages, in spite of their superficial differences. It is the
only LOGICAL CONSTANT, 'the one and only general primitive sign in logic',
since all logical operations, and hence all logical propositions and inferences,
are given with the very idea of a BIPOLAR elementary proposition (TLP
4.001, 5.47ff.; NB 22.1./5.5.15, 2.8.16).
'The general propositional form is: Things are thus-and-so (Es verhdlt sich
so und so)' (TIP 4.5). This formula is not restricted to true propositions. The
general propositional form is the colloquial equivalent of a 'propositional
variable' (TLP 4.53; OL 27, 30). It is the most general propositional vari­
able, the one which corresponds to the 'formal concept' of a proposition (see
SAYING/SHOWING). Its range of values is not a particular type of proposition -
'fa', fb', etc. - but the totality of propositions. Far from being vacuous, the
formula indicates that propositions must be logically articulate (composed of
function and argument), and must depict possible states of affairs, that is, be
descriptive.
In the first instance, the formula applies to ELEMENTARY PROPOSITIONS. But
according to the Tractatus, all propositions derive their representational char­
acter from such elementary propositions. The totality of propositions is
determined by the totality of elementary propositions (TIP 4.51-5.01; RUL
8.12). An essential part of the doctrine of the general propositional form is
the diesis of extensmnality: 'A proposition is a truth-function of elementary pro­
positions. (An elementary proposition is a truth-function of itself)' (TIP 5).
140
GENERAL PROPOSITIONAL FORM
The parenthesis alludes to the fact that in the TRUTH-TABLE notation every
elementary proposition is expressed as a truth-function (conjunction) of itself
and a tautology involving all other elementary propositions (e.g., '/>. (q v ~q),
etc.'). The first clause states that the truth of any proposition is dependent
analysed: 'propositions occur in other propositions only as the bases of
truth-operations'' (TIP 5.54). Consequently, the Tractatus has to explain away
the numerous intensional contexts of natural languages, such as the embedding
of a proposition in the scope of an intentional verb (in indirect speech or
ascriptions of propositional attitudes), causal explanations, scientific laws and
modal propositions. It does so with varying degrees of implausibility, either
by reducing these occurrences to extensional ones, as in the case of CAUSAL
explanations and ascriptions of BELIEF, or by denying that they constitute
genuine propositions, as in the case Of SCIENTIFIC laws and modal proposi­
tions (see LOGIC).
As a result of extensionalism, Tractatus 6 equates the general propositional
form with the 'general form of a trath-function': [/>, JV(^)]. This formula
specifies a series of propositions (the values of the most comprehensive pro-
positional variable) not through simply listing them (as in '/>, q, r') or through
a propositional function, as in the case of the quantifiers, but through a
'formal series' the members of which are ordered by a 'formal law' which
establishes internal relations between them (TLP 4.1252, 5.501). It does so
through a reiterable operation (TLP 5.23-5.3), one which can be applied to
its own results (TIP 5.251). Applying 0 to a one gets O'a, by repeating the
operation one gets O'O'a, and so on - as in 2, 4 (2 + 2), 6 (2 + 2 + 2), etc.
Such a series is determined by its first member and by the operation that
produces the following term out of the preceding one (TIP 4.1273). It is
expressed as [a, x, O'a] — a is the first term, x an arbitrary term, O'a the
form of x's immediate successor.
Truth-functional operations are reiterable, and hence create a formal
series expressed through an analogous ordered triple, p is the first member.
It is not a truth-function of elementary propositions (p. q.r'), but a list of all
elementary propositions (p, q, r, etc.). '£,' signifies not a random selection of
propositions, as Russell's Introduction claims, but a set of propositions which
has been constructed out of the initial set, and which may include elemen­
tary as well as molecular propositions (' ' indicates not generality, but that £
'represents', that is, lists, all of its values - TLP 5.501). JV(£) is the result of
applying the operation of joint negation to The operation jVis a general­
ized version of the diadic truth-operator 'neither p nor q', known as the
Sheffer stroke 'p j, q', from which it differs in that it operates on an arbitrary
number of propositions (this is important since the Tractatus is agnostic about
the number of elementary propositions) to yield a single proposition, the
joint denial of them all. In the truth-table notation the Sheffer stroke is
141
 GENERAL PROPOSITIONAL FORM
expressed as '(FFFT)(/>,a)', JV as '( T)(....)', with the right-hand brackets
indicating an arbitrary selection of n propositions, and the left-hand brackets
a truth-table with 2" — 1 F's omitted (TLP 5.5). In other words, the result of
applying JV to £ is true only if all the members of £ are false. For example,
i
'JV(p,q,r)' is equivalent to ~p.~q."~r'.
By specifying the general propositional form in this way, Wittgenstein
espouses a constructivist claim. Every proposition is 'a result of successive
applications to elementary propositions of the operation JV(£)' (TIP 6.001;
see NL 94, 102-3). The truth-functional connectives employed by Frege and
Russell - ' D ' , '•') ' V — are not just interdefinable, but can all be
reduced to the Sheffer stroke, and hence to JV (TIP 5.1311, 5.42, 5.441).
(M, a generalized version of the other Sheffer stroke 'p f q', namely 'either
not p or not q, etc.', serves the Tractatus's purposes equally well - RUL
19.8.19.) Sheffer had shown that 'J,' is 'expressively adequate' for the propo­
sitional calculus, capable of expressing all truth-functions of arbitrarily many
truth-arguments. The same holds, a fortiori, for the generalized version JV.
For example, all 16 truth-functions of two elementary propositions, the start­
ing-point of a formal series, can be generated through a process which starts
as follows: 1. JV(p,q) [~{p V q)], 2. JV(JV(p,q)) [<J> V q)], 3. JV(JV(p,q), JV(JV(p,q))) While such an expressively adequate notation is not explicit in the Tracta­
[~(~(p vq)v(pv q))] (contradiction), 4. _ JV(JV(N(p,q), JV(JV(p,q)))) tus, it is compatible with what Tractatus 5.501 says about the stipulation of
[—(-(/? v q) V [p V q))] (tautology). However, if we apply JV to that last
result, or even to any combination of the previous results, no new truth-
function is created. To get any further we have to bring into play a new
proposition, by applying JV to '/>' and 'o' individually, then to the resulting
'~P', ' ~ j ' ; and so on. This suggests that Wittgenstein's procedure does not
result in a formal series: it generates all the truth-functions of '/>' and 'q', but
not in a single definite order. Furthermore, if the initial set of elementary
N
propositions is infinite, Cantor's diagonal proof that 2 ° >No promises to
specify an infinite subset of elementary propositions which cannot be gener­
ated through Wittgenstein's procedure (but note that Wittgenstein rejects
Cantor's proof; see NUMBERS).
The operation JV is stretched even further by the task of constructing all
general propositions. Like propositional logic, quantification theory results
from the application of the same truth-functional operation. It differs merely
in the way in which the base of JV is specified, namely by a propositional
function. If £ has as its members all the values of the propositional function
f
fx, that is, the set of propositions fa,fb,fc, etc', then JV(£)' is the joint denial
of all these propositions, 'JV(/*)', and hence equivalent to '-(3*)/*'. If we
apply JV to that result, we get '(3x)/*'. If £ has as its members all the values
of JVifx), then 'N{JVifx))' is \x\fx', and 'JV{JV{JV(fx)))' is '~(x]fx'. This preserves
the unity of propositional and predicate logic: '-/>', '~{pv q)' and '~(3x]fic' are
c l
all expressed through the same operation, as JV{p)', JV(p,q)', 'JVifx)'.
It has been alleged that this notation is expressively inadequate, incapable
142
GENERAL PROPOSITIONAL FORM
of generating all formulae of first-order quantification theory. This inade­
quacy emerges in the case of multiply general formulae such as '(3x)(y)/x))' in
which the variables of the propositional function are bound by distinct
quantifiers. Thus, formulae are generated from the propositional variable
fxy', which has as its values fad, fab', fac', fba', etc. Applying JV to this
delivers the joint denial of these propositions, 'JVifxy)', which is equivalent to
l
''(^x^yYfxy', and applying JV to that proposition delivers N{Nifxy))', that is,
\3x)(ly]ficy'. Unfortunately, any further application of JV will simply take us
back and forth between equivalents of these two formulae (a similar dead­
lock results if we start from the propositional variable ~fxy). The problem is
not due to the employment of a single operator JV. 'Mfxyf is equivalent to
l
~{x){y]fxy', ^(Mifxy))' to \x){y)fxy', but then the flip-flop starts again. What is
needed is rather a way of operating separately on the two argument-places
of the propositional function. This can be done through enriching the JV-
operator by a variable-binding device. Thus, if one employs 'JVx(fx)' to indi­
cate the joint denial of all propositions resulting from substituting names for
the variable x, the previous stunibling-block '(^(ylfxy' can be expressed as
i i
N{Mx(Ny[Nfxy)))\ which translates as ~(x)~{y)~~fxy'.
values for a propositional variable, however, it expresses '(3x)/*' as
l
<N{JVx(fx))' = 'JV{JV(fa,fb,fc, etc.))', and \x)fx' as JVx(JV(fx))' = 'JV(JV(/a), JVifb),
JV(fc), etc.)'. In the first case, a single operation JV is applied to a possibly
infinite class of propositions, and then JV again to the result. In the second,
JV is applied to each of the members of that class, and then JV again to the
result. Now, it seems that this final operation presupposes the prior execu­
tion of (possibly) infinitely many steps, and has no immediate predecessor.
This would infringe the Tractatus's demand that all truth-functions be 'results
of the successive application of a finite number of truth-operations' (TLP
5.32). This objection is held to be on a par with the point that the truth-
tabular decision procedure cannot be applied to quantification over infinite
domains. But Wittgenstein was well aware of this, and his treatment of GEN­
ERALITY precisely avoids the need to go through an infinite number of steps,
by specifying the bases through propositional functions. And the original
objection ignores that what counts are stages of truth-functional construc­
tion, of which there are just two in our case: starting with an elementary
proposition fa' we apply JV once to yield 'JVifa)', which in turn delivers the
propositional function JVifx). Applying JV to all values of this function yields
'JV*(JV(/*))\
Although the Tractatus's constructivism is not obviously inadequate, it is
open to other objections. It does not provide the unitary explanation the
Tractatus sought, since the modified JV-operator occurs both with and without
a variable, as in 'JV(JVx(fx))'. Wittgenstein himself later pointed out that speci-
143
GENERAL PROPOSITIONAL FORM
fying the base of operations through a propositional function means some­
thing entirely different in the case of infinite domains. In that case, the 'and
so on' which is equivalent to the notion of successive operations cannot be
replaced by a list but indicates a new type of operation. To reduce logic to
a single operation is both impossible and superfluous. Since logical constants
do not represent logical objects anyway, we do not need to reduce their
number to niinimize our ontological commitments. The Tractatus's motive
for the reduction is the attempt, following Frege, to avoid piecemeal defini­
tions of the logical constants (TLP 5.45f., 5.46; NL 99, 105; Laws I §33, II
§§56-67, 139-44): in order to define them as they apply not just to elemen­
tary propositions (e.g., in '-/>'), but also to propositions which already con­
l
tain constants ('"(/>•?)', (3x)~fx'), one would have to introduce one
connective in advance of the others, with the result that one could not draw
on the latter in defining the former. But this problem can be avoided either
through recursive definitions, as in model theory, or by granting, as does the
later Wittgenstein, that signs like '~' and '.' may operate in slighdy different,
albeit analogous, ways in different areas.
Wittgenstein's later self-criticism focused not on his logical constructivism,
but on the very idea of a general propositional form. 'This is how tilings are'
looks like the general propositional form because it is used as a sentence-
schema which picks up the reference to an antecedent sentence. Although it
does not itself say something true or false, it sounds like an English sentence,
it consists of subject and predicate. This indicates that our concept of a pro­
position is in one sense determined by a Satzklang, and hence by the rules of
sentence-formation in a given language. But this does not provide the logico-
semantic essence sought by the Tractatus. And the idea that all propositions
say that such-and-such is the case is only a confused way of saying that a
proposition is whatever is true or false (expresses a thought/fact, can be the
object of a propositional attitude), that is, is an argument of the truth-func­
tional calculus (PLP 288-98, 372). But Wittgenstein's redundancy theory of
TRUTH implies that this provides neither a metaphysical insight into the
essence of propositions nor an independent test of propositionhood, since our
concept of truth and falsehood is itself part and parcel of our concept of a
proposition (PI §136; PG 124). To be sure, whether a sign-combination
counts as a proposition is a matter of grammar, not experience, but the rules
governing the term 'proposition' are neither sharp nor inflexible.
Wittgenstein also came to reject the thesis of extensionality, noting that it
characterizes the propositional calculus, but not ordinary language. More­
over, the idea that all propositions are, or contain, a description is part of the
misleading AUGUSTINIAN PICTURE OF LANGUAGE. Questions, commands, threats,
warnings and exhortations are Satze (this term, unlike 'proposition', is not
confined to indicative sentences). And many propositions, notably GRAMMA­
TICAL propositions, MATHEMATICAL propositions, and AVOWALS, have the form of
144
GENERALITY
indicative sentences, but, Wittgenstein argues, a non-descriptive role. PhUoso-
phical Investigations intimates that a PROPOSITION could be explained as a move
in the language-game. But even that explanation fails to cover tautologies
and contradictions (which don't say anything). For such reasons, the Investiga­
tions also rejects the very quest for a definition of 'proposition'. The concept
of a proposition is a FAMILY-RESEMBLANCE concept. It is explained through
examples. What holds together questions, commands, observation sentences,
first- and third-person psychological propositions, logical propositions, mathe­
matical equations, laws of physics, is not a single common essence, but a web
of similarities and analogies. It is possible to define sharply one type of propo­
sition, for example, tautologies. Indeed, the doctrine of the general proposi­
tional form defines the propositions of the propositional calculus as formulae
which are either bases or results of truth-functional operations. But this no
more shows that 'proposition' is not a family-resemblance concept, than the
fact that one can give sharp definitions of particular types of numbers shows
that 'number' is not a family-resemblance concept.
generality Frege's invention of a notation for quantification was crucial
to the development of modern logic. It allowed for the formalization of
statements involving multiple generality, which are essential to mathematical
definitions (e.g., of a continuous function) and proofs (e.g., of number-
theoretic theorems) (Notation §§11-12; Laws I §§8, 21-2). Frege analysed 'All
Greeks are bald' not into a subject 'all Greeks' and a predicate 'are bald',
but into a complex one-place function-name, 'if x is a Greek, then x is bald',
which is bound by the universal quantifier 'For all x'. Just as '* is a Greek'
is the name of a first-level truth-function mapping objects onto truth-values
(the True for Socrates, the False for Caesar), so this quantifier is a variable-
binding, variable-indexed 'second-level function' which maps first-level func­
tions onto truth-values — thus the universal quantifier has the value F for the
argument '* is a Greek' (for not everything is a Greek), T for 'x = x' (every­
thing is identical with itself). A 'general' proposition asserts that a function
0(x) has the value T for all arguments. 'Existential' propositions are not gen­
eral in this sense, but negations of general propositions. Thus, 'Some Greeks
are bald' comes out as 'Not for all x, if x is a Greek, x is not bald.' But like
'all', 'some' (which corresponds to existence) is a second-level concept, the one
which has 'falling within it' all first-level concepts which have at least one
object 'falling under' them (Foundations §53; 'Function' 26-7; 'Concept' 199-
202). Russell followed a similar line. He treated existence as 'a property of a
propositional function', but replaced Frege's cumbersome notation: \x)fx'
means that the propositional function fx is true in all instances, \3x)fx' that
it is true in at least one instance. The quantifiers, like the propositional con­
nectives, are names of 'logical constants', objects of which we have logical
experience (Principia *9; Logic 228-41; External 64-7; Introduction ch. XV).
145
GENERALITY
Wittgenstein praises the variable-binding apparatus '(*)' and '(3x)' for
having the 'mathematical multiplicity' required to express generality (TLP
4.04ff.; NB 23.10.14). It signifies what is being generalized, by showing
which part of its scope varies and which is constant, thereby distinguishing
between, for example, '(x)fx' (quantification over individuals) and '(<&)<!>#'
(quantification over properties). It has a determinate scope, which allows one
to distinguish, for example, truth functions of general propositions like
'(x)fxZ)(x)gx' from general propositions involving a complex function like
'(*)(/* D£*)'- Finally, it allows for one variable to fall within the scope of
another, which makes it possible to express multiple generality, and to dis­
tinguish '(x)(3j)xRy (e.g., 'For every natural number there is a greater one')
from 'fiy)(x)xRy' (e.g., 'There is a natural number which is greater than all
others').
At the same time, Wittgenstein rejects Russell's assimilation of \3x)fx' to
fx is possible', because the former can be false even if sentences of the form
'fx' express a logical (and indeed an empirical) possibility (TIP 4.464, 5.525).
More importandy, he attacks the idea that quantifiers are NAMES of LOGICAL
CONSTANTS, logical entities of some kind. The quantifiers no more stand for
second-level concepts or properties than '.' or ' v ' for relations between pro­
positions. Moreover, the arguments of quantifiers, what they operate on,
cannot be names (of first-level concepts), since they must be capable of
being true or false. This is shown by the fact that '(x)fx' can be negated not
just externally, '~(x)fx', but also internally, '(x)~fx\ a point highlighted by
Wittgenstein's T / F notation (see TRUTH-TABLES). Frege and Russell fail to do
justice to the fact that the understanding of general (i.e., universal) or exis­
tential propositions presupposes an understanding of ELEMENTARY PROPOSI­
TIONS, since the sense of the former is a function of that of certain
elementary propositions, and hence has to be explained by reference to
them (TLP 4.411; NL 106).
\x)fx' and '(^x)fx' express 'truth-functions', but those functions are not
things of any kind, but operations, namely of forming a logical product or
sum. The quantifiers differ from propositional connectives solely in the way
in which the base of the operation is specified, namely not through listing
these propositions, as in '(p. q) V r', but through a 'propositional variable' -
Russell's propositional function - fx'. Such a variable is a 'logical prototype';
it collects all the propositions of a certain form, since its values are all those
propositions we get by substituting a name for the variable, that is, fa, fb, fc,
etc. (TIP 3.315-3.317, 5.501, 5.522). Russell maintained that the Tractatus
'derives' \x)fi? from the logical product of its instances, fa .fb .fc...', and
'(Bx)fx' from the logical sum, favfb wfc...' ('Introduction' xv-xvi). And
Ramsey held that Wittgenstein's approach explains why fa' entails \3x)fx'
and \x)fx' fa', while Frege's explanation of '(3*1/*' as / h a s application' - a
statement of the form A(f) - does not (Mathematics \52>-A). Ironically, Tracta-
146
GENERALITY
tus 5.521 blames Frege and Russell for adopting this very approach (neither
Frege nor Russell did so explicity, although Principia I *59 comes close), and
Philosophical Grammar blames the early Wittgenstein (PG 268). The riddle's
solution is that the Tractatus attacks only the way in which Frege and Russell
linked generality to logical sum and product; while the later Wittgenstein
questions this link itself.
The Tractatus makes two points against Frege and Russell (TLP 5.1311,
5.523; NB 3.11./24.11.14, 2.12.16). (a) They confused two facts about the
universal proposition (x)fx: that it is a truth-function of all of its instances,
and that it is true just in case all the members of that set are true. The
former is the generality which (x)fx shares with (3x)fx, namely that it is a
truth-functional operation on all the values of a propositional function (fa,
fb, fc, etc.). It is expressed not by the quantifiers but by the 'argument', that
is, by the pattern 'x fx', while '()' and '(3)' differentiate the truth-functional
operations performed on the values of the propositional function, (b) They
derived the quantifiers from conjunctions and disjunctions. This is inadequate,
since the conjuncts/disjuncts involved are themselves instances (values) of
the propositional function; they already contain generality - fa' is equivalent
to (3x)(fx. x= a)' (TLP 5.47) - and hence cannot be used to explain it.
Moreover, they leave the relationship between general propositions and
their instances unexplained.
By clarifying that relation, Wittgenstein avoids some difficulties which
confront the straightforward identification of '(x)fx' with fa.fb.fc...' and of
\Bx)fx' with fay fb V fc' The first is that one can understand a general
proposition without ever having heard of a, b or c. Many Americans believe
that all communists are evil, without being able to name a single commu­
nist. 'The world can be completely described by means of fully generalized
propositions' (TLP 5.526; NB 17./19.10.14; PG 203-4). Nevertheless, a con­
nection with a logical product remains: some statement of the form
fa.fb.fc...' must be equivalent to the universal proposition, and this is why
I know that, for example, '~fa' is incompatible with \x)fx', whether or not I
have heard of a. Wittgenstein's account does justice to both points, since it
specifies the propositions involved through a propositional variable, and not
by enumerating them, which would require the use of particular names.
This also avoids a second problem. One can analyse '(x)fx' into a specific
conjunction fa\.fa ...
2 fa ' only if the number of objects in the universe is
n
finite. Even quantifying over a finite domain, for example, 'Everything in
this room is radioactive', is equivalent to a specific logical product like 'The
cup is radioactive. the table is radioactive' only if one adds the rider 'and
there is nothing else in the room', which in turn can be expressed as a spe­
cific logical product only if the universe does not contain an infinite number
of objects. This would make the explanation of quantification depend on an
'axiom of finitude', and is incompatible with the Tractatus's insistence that
147
 GENERALITY
there are 'no privileged numbers in logic': it is an empirical matter, settled
by the 'application of logic', how many objects there are (TLP 4.128,
4.2211, 5.453, 5.553). The Tractatus avoids that problem since it does not list
the elementary propositions which form the basis of the truth-functional
operation, and therefore does not use a specific list of names. However, this
in turn creates a problem detected by Ramsey (Mathematics 59—60, 153-5):
since completely generalized propositions do not refer to any specific objects,
it seems that they might allow more play to the facts than do the totality of
elementary propositions - contrary to Tractatus 5.5262. If the world only
contains a finite number of objects, we could construct a general proposition
which contains more distinct variables than there are objects. Take the
statement
(1) There are at least three individuals with some property.
In Russellian notation this comes out as
(1') (3x)(3y)(3z)(l<!>)(<l>x .<fy.^z.x¥=y.x¥=z.y¥^z).
But in a world in which there are only two individuals, Wittgenstein's
account of generality seems to turn (1') into a contradiction. If we substitute
f for '<!>', 'a' for 'x" and 'b' for y and 'z' (in our model-world no other
individual constant is available), we get
(1*) fa.fb.fb.a¥=b.a^b.b¥=b
a contradiction because of the last conjunct (by the same token, 'There is at
least one individual' ,and 'There are at least two individuals' come out as
tautologies). Ramsey's solution was to accept this consequence: statements
about the number of objects in the world are either tautologies or contra­
dictions. The Tractatus, by contrast, classifies such propositions as nonsensical
(TLP 4.1272, 5.535; NB 9.10.14). It can do so because it forgoes the use of
the LDENnTY-sign and analyses (1) as
(1") (3x)(By)(3 )m(®x-<b-®z).
Z a rule of construction.
Substitution into that proposition will not yield a contradiction. Nevertheless,
our commitment to objects in fully general propositions cannot outstrip the
number of objects, because the propositional function Ox will not have
values other than fa and fb, that is, there will be no other propositions for
truth-functions to operate on, which means that propositions employing
more variables than x and y have no application and are hence meaningless.
The number of objects in the universe, which Russell's axiom of infinity
tried to state, will be shown by the number of names employed by an ideal
notation, which in turn determines the number of variables that can be
meaningfully introduced.
Wittgenstein later detected flaws in the idea that, since \x)fx' entails 'fa', it
148
GENERALITY
must amount to the simultaneous assertion of all propositions of the form
fx'. That idea assumes that there is a totality of propositions of that form
which is both well-defined and, although not actually enumerated, capable
of being enumerated. Both assumptions, he later claimed, are unjustified.
(a) The Tractatus has been criticized for failing to realize that the truth-tabu­
lar decision procedure cannot be extended to the predicate calculus, since it
cannot be applied to infinite logical sums or products. But in fact the Tracta­
tus explicidy restricts that procedure to cases 'in which no sign of generality'
occurs (TLP 6.1203). However, it does assume that the logical operations
apply in the same way to a list of propositions and to the values of proposi­
tional functions. This holds for cases in which the class defined by the pro-
positional function is closed, such as the primary colours or the tones of the
octave. Here a list of names can be given in response to the question ' Which
objects are f?', and this list need not be followed by a cut-off clause, 'and
these are all', since, for example, the idea of a fifth primary colour is exclu­
ded as nonsensical by our grammar. 'In this picture I see all the primary
colours' is shorthand for 'I see red. I see green. I see blue...' The dots here
are 'dots of laziness' (LWL 15-16, 89-90; AWL 5-6; M 88-90; PR 117; PG
268-88; PI §208).
However, this does not hold for infinite classes: no matter how far the con­
junction '2 is even. 4 is even. 6 is even...' is extended, it never captures the
sense of the arithmetical 'all', which is given not through an enumeration or
any other description, but through a rule of construction, namely mathema­
tical induction (WVC 45, 51-3, 82; LWL 13-14; PR 150-1, 193-205; PG
432; BB 95-8). As a result, Wittgenstein, like the finitists (his pupil Good-
stein and Skolem), refuses to express general claims about infinite domains
through Russellian quantifiers and maintains that in such domains we
cannot state that there is an x which is / without stating a rule for specifying
which x is f. By the same token, universal statements about such domains
are never true accidentally (as 'All men are mortal'), but always by virtue of
The Tractatus's treatment is also inapplicable in cases in which the number
of possibilities is not so much infinite as indeterminate. The Tractatus had insis­
ted that while a proposition can leave something undetermined, it cannot be
an incomplete picture: what it leaves open must be specified, by the pre­
sence of a disjunction of possibilities (TLP 5.156; NB 16.6.15).
(2) There is a circle in this square
leaves open how the circle is placed within the square, but does so by speci­
fying that it occupies one of all the possible positions within the square.
Wittgenstein later saw that this is mistaken, not only because there is no
definite number of positions, but because the different positions 'are not
149
GRAMMAR
mentioned at all'. We do not verify (2) by going through every point. (2) is
best expressed through unbound variables which make clear that it does not
talk about a totality of possibilities, but is indeterminate (WVC 38-41; PG
257-67).
(b) The Tractatus's account implies that a logical sum and product can be
formed on any propositional function. However, once it is recognized that
the elementary propositions of form fx may not be logically independent, the
possibility of one type of operation no longer guarantees that of another
type: thus, it makes sense to say '(3x)(* is the colour of A)' but not '(*)(* is
l
the colour of A)', since A is red' is incompatible with 'A is green', etc.
Equally, while '(3x)(x is a circle. x is within the square)' makes sense, '(*)(* is
a circle. * is within the square)' does not. More generally, what makes sense
in some cases need not make sense in all cases. Wittgenstein diagnosed that
the Tractatus's account of generality was based on the mistaken view that
somehow any quantified proposition could be analysed into some logical
product or sum. But his critique is not confined to the Tractatus. The quanti­
fiers of the predicate calculus suggest that expressions of generality are topic-
neutral. Wittgenstein, by contrast, now claims that every propositional
system has its own rules of generality, and of the entailment between '(*)>'
and fa'. 'There are as many different "alls"' as there are types of proposi­
tions (PG 269; see LFM 262-70; LPP 51).
grammar According to the early Wittgenstein, 'distrust of grammar is the
first requisite of philosophizing' (NL 106). For the 'school-grammatical' form
of propositions disguises their LOGICAL FORM. The latter is revealed by an
ideal notation which follows the rules of LOGICAL SYNTAX, or 'logical gram­
mar' (TLP 3.325; see Logic 185, 269). Subsequendy, Wittgenstein used the
term 'grammar' to denote both the constitutive rules of language and the
philosophical investigation or tabulation of these rules (WVC 184; LWL 46-
8; BT 437; PI §90). Throughout his career he continued to use the term
'logic' or 'logic of language' (Sprachlogik) in the same two capacities (PI §§38,
90-3, 345; RPP I §1050; LW I §256; OC §§56, 501, 628), on the under­
standing that logical questions are really grammatical (Z §590), that is, con­
cern rules for the use of words.
Wittgenstein also speaks of 'the grammar of particular words, expres­
sions, phrases, propositions/sentences, and even of the grammar of states
and processes (BB 24; PI 18n, §187 - BB 1, 109; PI §660 - BB 70 - BB
51-3; PI §353 - PI §572; PG 82). But properly speaking, it is the corre­
sponding linguistic expression which has a grammar, namely a certain way
of being used. 'Grammatical rules' are standards for the correct use of an
expression which 'determine' its meaning: to give the meaning of a word is
to specify its grammar (M 51; PG 6 2 ^ ; OC §§61-2; LWL 34-9). 'Correct'
150
GRAMMAR
here does not mean 'true', since one can use a term in accordance with Lin­
guistic rules without saying something true. But a term is not used mean­
ingfully if it is applied to objects to which it could not apply truly. Equally,
to deny that a term applies to paradigmatic instances in a clear situation is
to display misunderstanding. The sense of a proposition is determined by its
place in the grammatical system, in that the latter determines its logical rela­
tions with other propositions (PG 152-3). The grammar of a language is the
overall system of grammatical rules, of the constitutive rules which define
that language by determining what it makes sense to say in it (PR 51; LWL
46-59; PG 60, 133, 143; PI §496). Unlike its predecessor logical syntax,
grammar is not universal - different languages have different grammars. But
the grammar of an individual word like 'understanding' is, in so far as other
languages have equivalent words.
The idea of grammar draws attention to the fact that speaking a language
is, among other things, to engage in a rule-guided activity. Some have
maintained that Wittgenstein did not subscribe to this normativist concep­
tion of language, and that his comparison of language to rule-following
activities is a misleading heuristic device, which betokens a school-masterly
attitude. This view may be motivated by his qualms about the idea of logi­
cal syntax as an arcane system of hidden rules. But Wittgenstein did not
abandon the idea that language is rule-governed, he clarified it, comparing
language no longer to a calculus but to a game (see LANGUAGE-GAME). Unlike
these analogies, the idea that language is rule-governed is not just a heuristic
device. Understanding a language involves mastery of techniques concerning
the application of rules (see RULE-FOLLOWING). And Wittgenstein continued to
stress the link between language, meaning and rules: 'following according to
the rule is FUNDAMENTAL to our language-game' (RFM 330; see BT
§45; OC §§61-2).
To assess Wittgenstein's normativist picture of language, one must
appreciate that his conception of rules is a functional one. Whether a sen­
tence expresses a grammatical rule depends on its role or function within
our linguistic practice. Wittgenstein distinguishes between 'empirical proposi­
tions' and 'grammatical propositions', sentences which are typfcally used to
express a rule (e.g. PI §§251, 458; AWL 31, 105-6; RFM 162). This distinc­
tion is not based on linguistic form - a grammatical proposition need not be
a meta-linguistic statement about how an expression is to be used. What
counts is whether we use it as a standard of linguistic correctness. The con­
trast between grammatical and empirical propositions is one between the
rules of our language-games, and moves in our language-games made in accor­
dance with these rules (e.g. PI §49; OC §622). The 'truth' of a grammatical
proposition consists not in stating how things are, but in accurately expres­
sing a rule. Grammatical propositions must be distinguished from empirical
statements to the effect that a community follows certain linguistic rules, for
151
GRAMMAR
example, 'All Englishmen use these signs in this way' (AWL 154; SDE 24),
and from propositions about the FRAMEWORK conditions which make rules
practicable. For they are used normatively, to explain, justify and criticize
uses of words.
Grammatical rules in this sense comprise not just school-grammatical or
syntactical rules, but also EXPLANATIONS of meaning (PG 68, 143; M 69; PR
78). What counts as an explanation of meaning is again a matter of func­
tion, not form. Such rules include (a) definitions, whether in formal
(' "Bachelor" means "unmarried man"') or material mode ('Bachelors are
unmarried men'); (b) analytical propositions ('All bachelors are unmarried');
(c) colour-charts and conversion-tables (BB 4; LFM 118); (d) OSTENSIVE DEFINI­
TIONS (BB 12, 90); (e) explanations by exemplification (PI §§69-79); (f)
expressions of the 'geometry' of colour like 'Nothing can be red and green
all over' (PR 51-2; LWL 8); (g) propositions of arithmetic and geometry
(WVC 38, 61-3, 156; PR 143, 170, 216, 249; LWL 8, 55; PG 319, 347;
RFM passim; see MATHEMATICS).
Wittgenstein's normativist conception of language contrasts sharply with
Quine's and Davidson's claim that the notion of a rule presupposes rather
than explains that of language. Less radically, one may protest that even in
Wittgenstein's extremely liberal sense rules are not prominent in our linguis­
tic practice. Here it is important to note that Wittgenstein did not insist that
rules are essential to learning a language (LW I §968). What counts is solely
whether our acquired practice can be characterized as rule-governed, where
this does not require that we actually consult the rules (LWL 48; PG 153;
PI §§82-3) but only that we could explain, criticize and justify our uses of
words by reference to rules (see CALCULUS MODEL), and do so when the occa­
sion demands it. Rules in Wittgenstein's liberal sense play a role in a host of
pedagogic and critical activities, some of which are institutionalized (educa­
tion, dictionaries): the teaching of language, the explanation of particular
words, the correction of mistakes, the justification of uses, the acquisition of
higher linguistic skills.
One might retort that this role is necessarily derivative. Grammatical rules
may distinguish between correct and incorrect use, but do not determine
either. For the rule is of no account unless it codifies existing practice, that
is, the prevailing practice. This reduces the difference between correct and
incorrect to that between conformity and nonconformity. Against this, Witt­
genstein righdy insisted that to use 'X' correcdy does not mean the same as
to use 'X' as most people do (RFM 406; Z §431). There is no incoherence
in the idea that a majority should commit linguistic mistakes (although in
most cases to use 'X" correcdy is to use it as most people do). Moreover,
although there may be grey areas, there is a difference between regularities
in linguistic behaviour and linguistic norms. While some deviations from
ordinary patterns are unusual (using 'sobriquet' instead of 'nick-name') or
152
GRAMMAR
inappropriate (referring to policemen as 'cops' in court), others are linguistic
mistakes - syntactic slips, malapropisms, wrong choices of words (but not all
of these amount to a meaningless use of words, as Wittgenstein sometimes
seems to suggest).
What is correct is that explicating and tabulating those grammatical rules
which are relevant to philosophy is not just secondary to linguistic practice,
but involves a 'one-sided perspective' on language (PG 68; BB 25), one
which emphasizes certarh features of our practice. Wittgenstein's PHILOSOPHY
adopts this perspective. It keeps the 'account book of language' and consists
of grammatical 'investigations', 'notes', 'analyses', 'remarks' or 'reminders'
(Erinnerungen) (PI §§89-90, 127, 199, 232, 392, 496, 574; PG 60). It reminds
us of the way we use words. One reason for doing this is that 'Essence is
expressed in grammar'; grammar determines 'what kind of object something
is' since it specifies what can be meaningfully said about it - 'Green is a
colour' is a grammatical proposition (PI §§371-3; PR 118; PG 463-4; BB
19; LSD 20). Empirical investigations into the physical nature of an object
or stuff X presuppose the grammar of 'X\ since the latter determines what
counts as X. The answer to the Socratic 'What is X?' is given not by
inspecting essences (abstract or mental objects), but by clarifying the mean­
l
ing of X", which is given by the rules for the use of 'X' (PI §383). More
generally, while metaphysics seeks to discover necessary truths about the
essential structure of reality, according to Wittgenstein the apparent struc­
ture of reality is nothing but a 'shadow' of grammar: he explains the special
status of logically necessary propositions through the idea that their role is
normative rather than descriptive. Grammar constitutes our FORM OF REPRE­
SENTATION, it lays down what counts as an intelligible description of reality,
and is hence not subject to empirical refutation.
Even if one accepts these claims, one may feel with Moore that Wittgen­
stein's liberal use of 'grammar' disguises the fact that philosophy must be
concerned with rules which are more fundamental than those of school-
grammar (M 69; LWL 97-8). Wittgenstein denies that there is such a differ­
ence between the two types of grammar. Both deal with rules for the use of
words, and there is no difference between 'contingent' and 'essential' rules,
'both lands of rules are rules in the same sense. It is just that some have
been the subject of philosophical discussions and some have not.' Philosophi­
cal grammar is special not in dealing with special rules, but in its aim,
namely of resolving philosophical problems. It differs from school-grammar
or linguistics merely in that (a) it is not concerned with exactness or compre­
hensiveness for its own sake; (b) it pays no attention to the history of lan­
guage or genetic problems in general; (c) its observations concern features
that are often shared by many languages (although they no longer concern
the logical structure of all possible symbolisms); (d) it has a wider, functional
conception of grammatical rules (AWL 31, 96-7; BT 413; PG 190; PI II
153
GRAMMAR
230; Z §§464-5). Indeed most of the rules which interest philosophy, for
example, 'One cannot know that p unless it is true that p\ do not concern
the grammarian, but that is because his purposes are not those of the philo­
sopher. However, one may grant the differences of purpose, while insisting
that they point towards fundamentally different kinds of rules. Thus Moore
insisted that philosophically relevant rules, like 'Nothing can be red and
green all over', exclude the inconceivable, while what children learn at
school, for example, 'You don't say "Three men was in the field" but
"Three men were in the field"', has nothing to do with philosophy. Report­
edly, Wittgenstein replied that this example indeed has nothing to do with
philosophy, since here all is perspicuous. But what about 'God the Father,
God the Son and God the Holy Ghost were in the field or was in the field?'
This rejoinder is inadequate. For we can easily distinguish the 'perspic­
uous' school-grammatical problem from the philosophical problem: the deci­
sion whether to employ the singular or the plural form would become trivial
if the problems concerning the Trinity could be intelhgibly resolved. It
remains tempting to hold that if the latter is a grammatical problem at all, it
concerns rules of a different kind (viz., concerning 'God' and 'person').
Nevertheless, one should grant Wittgenstein that there is a spectrum of
grammatical rules ranging from the philosophically insignificant ('The words
"north-east of" must be followed by a noun-phrase in the accusative') over
borderline cases ('The words "north-east of" must be followed by a terres­
trial place designation except "North Pole" or "South Pole"') to philosophi­
cally relevant cases ('The words "it is true that..." should not be used with
an adverb of time') (PIP 135-7).
Wittgenstein himself distinguishes between the 'depth grammar' and 'sur­
face grammar' of words (PI §664). Traditional philosophy goes wrong by
focusing on the latter, that is, the immediately evident (auditory or visual) fea­
tures of words, at the expense of their overall USE, which is like classifying
clouds by their shape (LC 2; AWL 46; PI §§10-14; Z §462). The surface
grammar (sentence structure) of 'I have a pain' is the same as that of 'I have
a pin', that of 'expectation' is that of a state (PI §§572-3) and that of 'to
mean' is that of an action verb like 'to say' (see PHILOSOPHICAL PSYCHOLOGY). But
their depth grammar is altogether different: the words have different combi­
natorial possibilities, and the propositions are different moves in the language-
game, with different logical relations and articulations. But it would be wrong
to think that this indicates a contrast between fundamentally different kinds of
grammatical rules, for example perspicuous ones violation of which produces
patent school-grammatical nonsense, and complex rules violation of which pro­
duces latent philosophical nonsense. Disregard for depth grammar yields
patent nonsense like 'I meant her passionately'; grammatical investigations
check whether philosophical positions lead to such nonsense (PI §464). The
metaphor of depth is misleading, since it suggests that depth grammar is dis-
154
GRAMMAR
covered through logical or linguistic analysis, as in the Tractatus or Chomsky.
The contrast is not between the surface and the 'geology' of expressions, but
between the local surroundings which can be taken in at a glance, and the
overall geography; that is, use of an expression. Moreover, Wittgenstein insis­
ted that (apart from being more difficult to recognize) most metaphysical pro­
positions are NONSENSE in the same sense as humdrum violations of grammar
and that grammar is flat, that is, that there are no METALOGICAL rules or con­
cepts which are logically more fundamental than others.
On Certainty explicidy raises the question of whether 'rule and empirical
proposition merge into one another' (OC §309). Three possible grounds for
a positive answer can be detected. One is that some propositions which
have the form of empirical propositions are among the hinges on which our
language-games turn. Another is that we can imagine circumstances in
which certain sentences turn from grammatical propositions into moves of
the language-game. Finally, 'there is no sharp boundary between proposi­
tions of logic and empirical propositions.' But Wittgenstein adds that this
'lack of sharpness is that of the boundary between rule and empirical propo­
sition' and due to the fact that the concept of a proposition is itself vague
(OC §§56, 82, 318-20, 622). That a division is not 'sharp' does not mean
that it is unworkable. Moreover, the first two possible grounds are accom­
modated in Wittgenstein's functional conception of grammatical rules: the
logical status of a sentence is due not to its linguistic form, but to the way it
is used, and can hence change: 'any empirical proposition can be trans­
formed into a postulate — and then becomes a norm of description' (OC
§321). It is true that Wittgenstein suspects this statement to be reminiscent of
T I P (in fact it recalls PR 59). But what he means is that it would be dog­
matic to insist that any proposition could change its logical role, since the
revisability of our form of representation is restricted. The famous metaphor
of the 'river-bed of thoughts' distinguishes between 'the movement of the
waters on the river-bed' (changes in empirical beliefs), 'the shift of the bed
itself (conceptual changes brought about by adopting new grammatical
rules), and the 'hard rock' of the river-bank which is not subject to altera­
tion (OC §§95-9). The last includes propositions of logic which partly define
what we mean by thinking, inferring, language, while the shifting sand of
the bank consists of propositions which we could use either normatively or
descriptively. The only significant concession of On Certainty to the view that
there is no boundary between empirical and grammatical propositions,
namely that even among empirical propositions some (e.g., 'The earth has
existed for over a hundred years') must be CERTAIN, is more damaging to
rationalism than to Wittgenstein's earlier views (OC §§401-2, 558; WAM
70—5). In contrast to necessary propositions, the reason is not that their
negation is excluded as nonsensical by a specific grammatical rule, but
rather that abandoning them would undermine our whole system of beliefs.
155

H Bodies Wittgenstein's dictum implies that a human being cannot be identi­
human being 'Only of a living human being and what resembles
(behaves like) a living human being can one say: it has sensations; it sees; is
blind; hears; is deaf; is conscious or unconscious' (PI §281, see §§282-7,
359-61). Wittgenstein is aware that we apply some psychological terms to
inanimate objects like dolls, but maintains that this is a 'secondary use',
since it involves endowing these objects with the behavioural capacities of
human beings, as in a fairy-tale. In Wittgenstein's work one can detect two
interrelated rationales for this famous claim. One is that there is a con­
ceptual rather than merely empirical connection between psychological
terms and certain forms of BEHAVIOUR. One can meaningfully ascribe
THOUGHTS only to creatures that are in principle capable of manifesting
them. Mutatis mutandis, sensation-names and perceptual verbs can be applied
only to sentient creatures who react to their environmerit and can display
pleasure and pain, that is, to animals, although this is a matter of degree (PI
§284). The second rationale is that such expressions make sense only as part
of the complex weave of a form of life: 'The concept of pain is character­
ized by its particular function in our life . . . we only call 'pain' what has this
position, these connections' (Z §§532-3). On one understanding, there is a
tension between the stress on behavioural CRITERIA and the FORM OF LIFE con­
textualism. For, the latter seems to imply that in a community of people
who display, say, pain-behaviour just as we do, but do not react to it with
sympathy, 'pain' would not have the same meaning. On the other hand, the
two are compatible if the context of a form of life is needed because without
it certain forms of expressive behaviour would be unintelligible (although
this does not hold of pain-behaviour).
If Wittgenstein's restriction of experiential predicates to sentient creatures
is correct, it makes no sense to ascribe psychological terms either to a non-
corporeal soul-substance, as Cartesianism does, or to the human body or
one of its parts, as materialist theories, in particular mind-brain identity the­
ories, do, or to a machine, notably a computer, as cognitive science and
functionalism do. Although most contemporary philosophers accept the first
corollary, they would maintain that the second two have been overtaken by
156
HUMAN BEING
science, which informs us that the mind is the brain, and which has con­
structed computers which are capable of playing chess or performing com­
plex calculations. However, the issue is far from straightforward.
fied with a body, but has a body. Saying 'Carter's body is in pain' instead of
'Carter is in pain' would amount to a shift in grammar (PI §283; BB 73).
This points in the direction of Strawson's claim that a human being or
person is neither a Cartesian soul nor a body, nor a composite of the two,
but a distinct type of thing to which both physical and mental predicates
apply, and which can be said to have rather than be a body. The conclu­
sion is supported by the fact that substituting body-referring expressions
('Carter's body') for personal names saha significatione and salva veritate is possi­
ble only for non-psychological predicates (as in 'Carter is sun-burnt all over',
but not in 'Carter intends to go to London' or 'Mary loves Carter').
Parts of the body We apply psychological expressions metaphorically or
metonymically to parts of the body, for example in 'She has a generous
heart' or 'My hand hurts.' But it is noteworthy that in the latter case we
comfort the sufferer, not the hand (PI §286; Z §§540-1). Equally, 'My brain
is numb' means 'I can't think.' But it is nonsense to say either of a person's
mind or of his brain that it has a toothache, sees the sunset, intends to go to
London. It is equally nonsensical to say, using the terms with their standard
meaning, that a brain classifies or compares, asks questions, and answers
them by constructing theories. Nothing a brain can intelligibly be said to do
(e.g., emitting electric impulses) could constitute thinking, etc. Such things
are regularly said about the brain by psychologists, neurophysiologists and
philosophers, and that use could be seen as a legitimate extension or techni­
cal use of psychological vocabulary, but it needs to be explained. This is
done by those who apply epistemic terms to the brain only in the purely
technical sense of communication theory. But often such applications surrep­
titiously trade on our non-technical use of epistemic terms, which are tied to
human behaviour, or they boil down to the claim that the processes which
are thought to license such applications are correlated with experiences as a
matter of empirical fact (thus Wittgenstein held that localizing thoughts in
parts of the brain uses the expression 'locality of thought' in a different
sense (BB 7)). If this is correct, explaining the mental capacities of human
beings through applying mental terms to the brain where they cannot have
their standard meaning is a homunculus fallacy, on a par with postulating a
little man within our brain.
Computers We speak of computers as calculating and playing chess. Witt­
genstein, by contrast, suggested that to say of a machine that it thinks is a
157
HUMAN BEING HUMAN BEING

category mistake, not because thinking is an occult process in a mental is possible only for a creature which possesses conation and a will, that is,
realm (BB 47; PG 106), but because manipulating symbols is by itself not can take an interest in things and pursue goals. None of this holds of a
sufficient for thinking or understanding (in this respect his reasoning paral­ computer, even if it passes the Turing test. However, this would not rule out
lels Searle's celebrated Chinese Room argument). Manipulating symbols ascribing psychological attributes to science-fictional android robots who
counts as calculating or inferring only if it is a normative rather than merely move about, react to their environment, talk and solve problems, feel plea­
mechanical activity. It must be possible to distinguish between a prediction sure and pain, etc. But such robots would no longer count as machines.
that such-and-such a result will be obtained and judging a result to be cor­
rect. While machines can act in accordance with rules, provided they are
properly programmed and functioning, only human beings (or creatures
who 'resemble' - behave like - them, which 'machines' as thought of here
do not) are capable of following rules, and hence of calculating, etc. For
only they are capable of justifying or correcting their proceeding by refer­
ence to the rule (see RULE-FOLLOWING).
This reasoning underlies Wittgenstein's remark: 'Turing's "machines".
These machines are humans who calculate' (RPP I §1096). Turing had sket­
ched the notion of a machine which would compute number-theoretic func­
tions by analogy with human beings. Wittgenstein's point is that what
Turing machines do counts as genuine calculating only if it is performed by
human beings. For although such human beings would presumably be cal­
culating mechanically, that is, without consulting rules, they would be cap­
able of calculating non-mechanically, they could invoke rules or recognize
mistakes (RFM 234, 257-8, 382, 422; LFM 36-9; RPP I §560).
A natural objection is that computers of a certain complexity can do the
same: when asked why they produce a certain result they cite the relevant
rules, and they could be programmed to commit and rectify mistakes. And
the problem about Turing machines being too mechanical for calculation
seems taken care of by the so-called 'Turing test', according to which whe­
ther or not a computer can trunk depends on whether or not the answers it
gives on a screen are indistinguishable from those a human being might
type out. Answers which could pass that test would have to be unpredictable
and non-wooden. The Turing test is in line with Wittgenstein (and in con­
flict with Searle) in so far as it decides on whether or not computers think
by reference not to their internal constitution (being made of silicon rather
than carbohydrates), but to their capacities, what they can do. But Wittgen­
stein's dictum implies that the test is mistaken to treat the appearance of
symbols on a screen as a form of behaviour which could display thought.
Typing out symbols counts as 'dunking' and 'calculating' only for a creature
to whom we can ascribe a wider range of psychological attributes (RPP I
§563). It is plausible to hold that neither experiences nor emotions can be
ascribed to a machine which merely reacts to the pushing of keys in accor­
dance with a programme. But this by itself does not preclude one's talking
here of thought or calculation. The ultimate rationale for Wittgenstein's
position is that rule-following requires the doing of things for a reason, which

158 159

I at its centre. When the despot has a toothache, he (and everybody else) does
I/self There are three traditional theories concerning the first-person pro­
noun. According to Cartesianism it refers to a soul-substance attached to the
body; according to Humean theories it can refer only to a bundle of mental
episodes, since no such unitary substance is encountered in introspection;
according to Kantianism it signifies the transcendental unity of appercep­
tion, a formal feature of all judgements, namely that they can be prefixed
by 'I think'. Wittgenstein imbibed these options through Schopenhauer's
Kantianism and Russell's Humeanism. Both rejected the Cartesian soul-sub­
stance, but retained various 'selves'. Schopenhauer turned the transcendental
unity of apperception into the idea that the subject of experience cannot
itself be experienced (World II ch. 41). Russell first thought of T as the logi­
cally proper NAME of a self we know by acquaintance, then as the mere
grammatical subject of psychological predication known by description
(Problems 27-8; 'Theory' 36-7).
'The I, the I is what is deeply mysterious' (NB 5.8.16). Wittgenstein trans­
posed the perplexities of Humeanism and Kantianism onto a linguistic
plane. Thus, the transcendental SOLIPSISM of the Tractatus involves removing
the first-person pronoun from the analysis of BELiEF-sentences like 'I think
that p\ The immediate reason is the Humean idea that the I is not found in
experience. But the deeper rationale is Kantian. Whatever we experience
could be otherwise. By contrast, it is a priori that the experience I now have
is my experience, or that this visual field is my visual field. Hence I cannot
ascribe experiences to myself in BIPOLAR propositions. The subject of experi­
ence does not just happen to elude introspection, it could not be encoun­
tered in experience, since its connection with experience is not a posteriori.
The idea that T can be eliminated from our language survived the transi­
tion from transcendental to methodological solipsism. A 'phenomenological'
language which refers to subjective experiences is semantically basic; but the
experiences of this language have no owner (see PRIVACY). Following Lichten-
berg, instead of 'I think' we should say 'It thinks', as in 'It rains' (WVC 49-
50; M 100-1; PR 88-90). The reason is not that 'thinking' is a feature-
placing predicate like 'raining', but that it is logically impossible for anybody
else to have what I have when I have a pain, since no one else could have
160
I/SELF
a pain which / could encounter. Accordingly, 'I have a toothache' is analy­
tic; it is senseless to say that / as opposed to someone else has the pain. In
first-person present tense psychological utterances T is redundant.
The eliminability of T is illustrated by a fictional language with a despot
not say
(1) I have (He has) a toothache
but simply
(1') There is a toothache.
However, when someone else - N.N. - has a toothache, everybody, includ­
ing N.N., says
(2) N.N. is behaving as the Centre behaves when there is a toothache.
Wittgenstein claims that this Lichtenbergian language makes perspicuous
what is essential to our form of representation. But as regards anybody
other than the Centre, this language differs sharply from ours: (a) it pre­
cludes others conjecturing that N.N. has a pain not manifested in his beha­
viour; (b) it makes N.N.'s report subject to error, since it is possible to
mischaracterize one's own behaviour; (c) because (2), unlike (1), is based on
behaviour, 'pain' is ambiguous; real pain could be attributed only to the
Centre. The methodological solipsist might respond that each one of us
employs a mono-centred language which is adequate for its centre. But even
for the centre such a language is inadequate. T occurs not just in proposi­
tions like (1), but also in introductions like 'I am N.N.' or propositions like 'I
am not N.N.', which cannot be reproduced in the mono-centred language.
Wittgenstein later tried to accommodate these points by abandoning the
simple elimination of T in favour of cfistmguishing between its use 'as sub­
ject' in propositions like (1) and its use 'as object' (BB 66-7; see M 100-3;
PR 86) in propositions like
(3) I have broken my arm.
Unlike (1), (3) is not immune to doubt or error. Accordingly ,T is either
essentially redundant - as in (1) - or refers to my body - as in (3).
Wittgenstein never came back to this 'dual-use' view, and it is flawed in
several respects. Although in uttering (3), I may be in doubt or error, this
cannot arise from misidentification. To be sure, in a rugby-scrum I might
mistake my arm for yours, but I do not misidentify myself, or mistake myself
for you. Moreover, the dual-use view implies that a first-person proposition
like
161
 I/SELF
(4) I am writing a letter
must be analysed into a proposition about bare bodily movements, plus an
infallible first-person experiential proposition, for example about volitions.
Later, Wittgenstein realized that this distorts both the subject and the pre­
dicate of such utterances. (4) does not ascribe a 'colourless' bodily move­
ment, but a human action which is infused with rather than accompanied
by intentions (see BEHAVIOUR AND BEHAVIOURISM). Equally, its subject is not a
body, but a HUMAN BEING. T and 'this body' are interchangeable saka veritate
in propositions like (3), but they are not even interchangeable saka sig-
nificatione in propositions like (4). This does not mean that T is ambiguous,
but simply that its function and that of 'this body' are analogous for part of
their range, but diverge elsewhere.
This rules out the materialist alternative to traditional accounts, according
to which T refers to a body or one of its parts, for example the brain. But
the fact that T does not refer to a body does not entail that it refers to an
entity (ego, soul, self) attached to the body. For Wittgenstein, it is never­
theless essential to our use of T that it is emitted by creatures that nave a
body - on the Kantian grounds that there are no criteria of identity for
soul-substances (AWL 24, 62; BB 69; LPE 300, 308). The obvious alter­
native is Strawson's suggestion that the first-person pronoun refers neither to
a body, nor to a self, nor to a bundle of mental episodes, but to a 'person',
a living creature with special mental capacities. Obviously, T is no more the
proper name of a person than 'here' is the name of a place (LPE 298; PI
§410). Equally, it does not mean the same as 'the person who is now speak­
ing', since it cannot be substituted saka veritate in 'The person who is now
speaking is the headmistress.' Nevertheless, it seems that its meaning is given
by the rule that T refers to the person that uses it. In spite of his focus on
human beings, Wittgenstein questioned this proposal, and occasionally flady
rejected it: 'It is correct, although paradoxical, to say: " T does not refer to
(bezeichnet) a person'" (MS 116 215; LPE 283).
This claim that T is not a referring expression cannot be defended by
reference to the idea that first-person psychological utterances are AVOWALS
rather than descriptions. Its use is not confined to avowals, and referring is
not tied to description ('God save the Queen!'). A possible defence is the
suggestion that 'I am N.N.' is not an identity statement, on the grounds that
it is not verified like ordinary identity statements ('This is Lewis Carroll'),
and is typically used to introduce oneself. However, 'I am N.N.' can be used
to affirm an identity (e.g., when I find out that I have been nicknamed
'N.N.'), which means that T remains a candidate for referring.
Wittgenstein himself focuses on a different line of argument, which devel­
ops a kernel of truth in his previous accounts (PI §§398-411; BT 523).
There are substantial differences between, on the one hand, the first-person
162
I/SELF
pronoun and, on the other, person-referring expressions such as proper
names, descriptions, personal pronouns, and demonstratives ('this', 'that').
(a) T does not allow of referential failure: (i) T cannot mischaracterize that
which the speaker intends to talk about; (ii) as we have seen, I cannot mis­
take myself for you, but only ascribe to myself something that applies to
you; (iii) when I utter 'The present King of France is bald', the grammatical
subject does not apply to anything, but there is no such risk in uttering T-
sentences. For these reasons, an amnesiac can use 'I' correcdy without
knowing whether he is speaking of N.N. or the Oer. What he needs to
know is only that he is speaking about himself, and this he is assured of by
virtue of knowing that it issues from his mouth. In contrast to other personal
pronouns, using T does not presuppose the possibility of identifying the
referent through a name or description, or through a deictic gesture.
However, instead of concluding that T is not a referring expression, one
might conclude that it is a j«/)«r-r«/^m/^expression, one guaranteed against
referential failure. One could compare ordinary referring to shooting an
arrow at a bull's-eye on a wall, which one may hit or miss. In these terms,
the use of T is analogous not to a magic arrow that always hits its target, as
a completely unambiguous set of names or definite descriptions might be,
but to drawing a bull's-eye around an arrow already stuck in the wall. The
idea of hit or miss has no place, and this marks a logical difference from
ordinary referring expressions.
(b) For the user himself, T does not identify someone,, in the sense of sin­
gling out someone from a group of people or things, although its use may
enable others to identify someone. 'I don't choose the mouth which says "I
have toothache'" (LPE 311; MS220 25). It might be objected that I single
myself out from others, for example, in confessing that / and no one else
broke the vase. But this is a case not of identifying oneself, but of drawing
attention to oneself; for it does not specify who it is - N.N., the Oer - that
broke the vase. One might reply that it does so in the same way in which
deictic expressions do. However, to say T is not thereby to point at any­
thing, it is more akin to raising one's arm. T signifies the point of origin of
the system of deixis, not a point on the deictic graph (BB 67-8; LSD 33;
BT 523).
Ultimately, the question of whether T is a referring expression depends
on what one makes of the term 'referring', an issue Wittgenstein did not suf-
ficiendy discuss. It seems clear that T, as well as 'you', 'she', etc., may be
used to refer to a single person. And in uttering 'I broke the vase' I tell you
who did it. Just like 'H.G.', T helps determine the sense of propositions in
which it occurs, and it does so by determining whom the proposition is
163
IDENnTY
about. It makes that contribution in a way which differs from that of other
referring expressions. But this might best be expressed by saying that T is a
degenerate case of a referring expression, just as tautologies are degenerate
cases of propositions.
identity (the same) This concept played an important part in the logicist
programme of reducing arithmetical equations to logical propositions, and
for this reason '=' was introduced into logic by Frege to denote a binary
truth-function governed by special axioms (Laws I §§4, 7, 47). This created a
puzzle. If statements like
(1) The morning star is the evening star
express a relation, is it one between objects, or between the names which
represent them? In 'Sense and Reference', Frege rejected the second option
(adopted in Notation §8), on the grounds that if (1) were about signs it would
not express proper knowledge, since the connection between a name and its
object is arbitrary. The first alternative, however, seems to imply that (1) is
equivalent to
(2) The morning star is the morning star
which is an instance of the traditional 'law of identity' - every object is
identical with itself. But unlike (2), (1) is informative, it expresses an empiri­
cal discovery. In response, Frege distinguished betweerr two aspects of an
expression's content, its 'MEANING', that is, the object it stands for, and its
'sense', the mode of presentation of its meaning. 'The morning star' and
'the evening star' mean the same object - Venus - but present it in differ­
ent ways, which is why (1) differs from (2).
Like Frege, Russell used '=' to formalize count-statements. While he
rejected the sense/meaning distinction, his theory of descriptions actually
elaborates Frege's gnomic solution to the puzzle of non-trivial identity state­
ments (Logic 39—56; Principia I 66-71). That NAMES like 'the morning star'
have sense can only mean that they are abbreviations of definite descrip­
tions, for example, 'the planet visible in the morning sky'. On this basis, (1)
can be analysed as making three substantial claims: there is exacdy one
planet visible in the morning sky; there is exacdy one planet visible in the
evening sky; whatever planet is visible in the morning sky is also visible in
the evening sky. Identity statements of the form 'a = the F, for example
(3) Scott is the author of Waverky
=
come out as '(3x)(/*. (y)(fy D * J>) • x — a)', or, more simply,
(3') fa.(?)(fyDy = a)
Wittgenstein followed- Frege and Russell in holding that 'is' and its cog-
164
IDENTITY
nates in natural language are ambiguous between identity (as in (1)), pre­
dication ('Socrates is mortal') and existence ('There is a God'). Throughout
his career, he diagnosed this as a cause of confusions such as the Hegelian
paradox of 'identity in difference' (TLP 3.323; RCL; LWL 4; PG 53; PI
§558) and suggested that these might be forestalled through a notation
which replaces 'is' respectively through '=', 'e' and '(3*)' (TS220 §99). This
is the only exemplification of a method envisaged in Philosophical Investigations
§90, namely of dissolving philosophical problems through a new notation.
But it is not in line with the early work, which bans '=' from the ideal nota­
tion as part of its flight from LOGICAL CONSTANTS. Like quantifiers and propo­
sitional connectives, '=' is already present in elementary propositions, since
'fa' says the same as \3x)(fx. x = a)' (TLP 5.47; NM 117). However, 'Identity
of object I express by identity of sign, and not by using a sign for identity.
Difference of objects I express by difference of signs' (TLP 5.53; NB
29.11.14). While ordinary language often employs several names for a single
object and the same name for different objects, an ideal notation uses a dif­
ferent sign, for every object. Just as the existence of an object is shown in
language by the use of a name (TLP 5.535), identity is shown by the use of
the same name. One motive for this strategy is the Tractatus's insistence that
all necessity is truth-functional (see LOGIC). Apparent necessary truths like
'a=a' or '(x)x = x' cannot be reduced to TAUTOLOGIES (RUL 17.10.13), and
hence must be treated as 'pseudo-propositions'. A possible justification for
this treatment runs like this:
Pi Any meaningful proposition can be understood without knowing
whether it is true (TLP 4.024).
P2 One cannot understand two names without knowing whether they
refer to the same object or to two different objects (TLP 4.243,
6.2322).
C If 'a' and 'b' are names, 'a = b' cannot be meaningfully expressed; like
'a = a' or '(x)x = x' it is a pseudo-proposition (TLP 5.534).
Even if one waives the suspicion that Pi, a consequence of the principle of
BIPOLARITY, is dogmatic, P obviously does not hold for ordinary singular
2
terms. It does hold, however, for the logically proper names postulated by
the Tractatus. In their case, 'a — b' is either necessarily true or necessarily
false. Since it is not molecular, it cannot be a tautology, and must hence be
nonsensical.
Another line of thought underlies the claim that 'to say of two things that
they are identical is nonsense, and to say of one thing that it is identical with
itself is to say nothing at all' (TLP 5.5303). According to Quine, this ignores
that identity statements can be true and non-trivial because different sing­
ular terms can refer to the same thing. In fact, Quine's point reinforces the
165
IDENTITY
Tractatus. Proposition (1) does not talk about signs, but the knowledge it
expresses is also expressed by
(1') 'The morning star' and 'the evening star' name one and the same
stellar body.
(1) only has a role because we can refer to a single object through different
singular terms. This role becomes obsolete in a symbolism in which every
thing has its own unique name: all propositions of the form 'a = V would be
incorrect, and those of the form 'a = d poindess (NB 5.-6.9./11.11.14). That
notation is superior because it avoids the suggestion that identity is a rela­
tion like 'x loves j ' , albeit one that every thing has to itself, but to no other
thing. That this is mistaken becomes clear if one looks at the statement
'Only a is / . The Russellian paraphrase of this is the second conjunct of
(3'), that is, '(y)(fy Z)y = «)'• But 'what this proposition says is simply that only
a satisfies the function/, and not that only things which have a certain rela­
tion to a satisfy the function / Of course, it might then be said that only a
did have this relation to a, but to express that, we should need the identity-
sign itseLf (TLP 5.5301). The last sentence draws attention to the fact that
Russell's analysis cannot clarify the nature of the alleged relation between a
and b in 'a = b\ since the problematic sign reoccurs in the second conjunct
of the analysans. That clause amounts to the claim that there is at most one
f which is best expressed by
(3*) ~(3x)(3y)(fx.ty.x*y).
No such paraphrase is available for a genuine relation like 'x loves y'.
Whereas the basic use of such relational expressions is in unquantified pro­
positions like 'a loves b\ '=' is explained by reference to its occurrence within
the scope of a quantifier. This suggests that the sign itself is part of the appa­
ratus of quantification, which is the point of Wittgenstein's analysis (NB
29.11.14, 12.5.15). Identity and difference of names indicates identity and
difference of objects - 'fab. ~ a = V is simply faV, fab .a = V simply fad (or
'jbb\ Equally, if one adopts Wittgenstein's convention for reading the quan­
tifiers, then identity and difference of variables indicates that the same or
different names must be substituted. Accordingly, (3') can be written as
(3#) fa. ~{3x,y)(fx.fy)
Other uses o f ' = ' can be treated along the same lines (TLP 5.531-5.533):
English Russell TLP
There are at least 2 things which a r e / (3x)(3y)(fx.jy. ~x-y) (Bx)[3y)fc .Jji
Somebody loves himself (3x)(3y)(fx.jy. x=y) (3*1/**
Somebody loves somebody (Bx)(3y)xRy (3x)(3y)xRy V (Bx)xRx
166
IDENTITY
Ifanythingis/itiso (#D'=«i M(/0»-~<MM/*-Jy)
There are at most 2 things which are / (x)(y)(z)((fx -jy •/<:) D
(x=y v x = ; v ; ^ ) ~(Bx)(By)(Bz)(fi.Jy.fz)
Precisely one thing i s / (*)W(/0*=.v) (3*1£• ~ $>$Mjx• fy
This removal of '=' has far-reaching consequences, (a) The Tractatus does
not deny that ordinary identity statements like (1) and (2) make sense. They
involve signs of complexes, which are analysed into a description of their
simple components. But this LOGICAL ANALYSIS does not employ '='; it indi­
cates the identity of simple objects through using the same name, (b) Fre­
ge's axioms governing identity are pseudo-propositions (TLP 5.534); at best
they try to say something which shows itself in the logical structure of
ordinary discourse, (c) The problems created by Russell's axiom of infinity
(the claim that there is an infinite number of objects in the universe) cannot
arise. For claims about how many objects there are, in contrast to claims
about how many objects of a particular kind F there are, cannot even be
formulated (see GENERALITY). By the same token, that a simple object a exists
cannot be expressed by \3x)(x = a)\ (d) Mathematical equations are pseudo-
propositions. They do not say anything about the world, but equate signs
which are equivalent by virtue of rules governing reiterable operations (TIP
6.2ff.).
Identity played a major role in Ramsey's attempt to improve the logicism
of Principia. He followed Wittgenstein's criticism of Russell's definition of
identity, namely that it implies the principle of the identity of indiscernibles,
that is, that two objects cannot have all their qualities in common (Principia
*13.01; T I P 5.5302; Mathematics 30-1). Unfortunately, like Russell and Witt­
genstein he ignored the question of whether these qualities are to include
spatio-temporal location, an inclusion which would make the principle plau­
sible. At the same time, Ramsey tried to retain identity in a way which
accommodates the Tractatus, arguing in effect that true identity statements
are tautologies and false ones contradictions. In reply, Wittgenstein insisted
that a false identity statement involving logically proper names is nonsensical
rather than contradictory, and that the same holds for true identity state­
ments, since the negation of a nonsense is itself a nonsense (RAL 2.7.26, 7./
8.26).
After abandoning the idea that only truth-functional necessity is expres­
sible, Wittgenstein allowed necessary propositions involving identity. How­
ever, he continued to deny that identity statements describe a peculiar
relation in which everything stands to itself. Although this is not evident
from their form, statements of the type 'a = V have the role of GRAMMATICAL
propositions: they express substitution rules which license transformations of
empirical propositions, for example from 'Oa' to 'OA', and exclude certain
167
IDENTITY
propositions as nonsensical, for example 'Oa.~OA'. Thus a mathematical
proposition like '12 x 12= 144' licenses the inference from "There were 12
rows of 12 chairs' to 'There were 144 chairs.' Equally, an identity statement
like (1) licenses the inference from 'The morning star is a planet' to 'The
evening star is a planet', and excludes 'The morning star is a planet but the
evening star isn't.' Such statements can be informative because given our
criteria for understanding singular terms it is possible for someone to know
what the morning star is, and what the evening star is, without knowing
that they are identical. The role, if not the form, of (1) is that of a rule for
the use of words. This casts doubt on Kripke's claim that (1) is a posteriori
but necessary: the discovery (1) expresses is a contingent fact, namely that a
single object satisfies two descriptions, or is the bearer of two names. What
could be said to be necessary, but would be a priori, is the identity of a
thing with itself.
Even that, however, presupposes the legitimacy of propositions of the
form 'a — d (the. law of identity), which Wittgenstein continued to question.
'a = a' looks like a genuine truth because its negation strikes us as obviously
false, and is the result of applying the substitution rule 'a = V to itself. But
like 'Everything is identical with itself or 'Everything fits into its own
shape', 'a = a' is degenerate. The 'partners' of the apparent relationship are
not independent. Wittgenstein maintains that it would make sense to say
that a is identical with itself only if it could fail to be so, which is impossible
since 'a # a' is a nonsense, and so is trying to distinguish something from
itself. Accordingly, 'a = a' is 'nonsense', a 'perfecdy useless proposition' (PI
§216; LFM 26-7, 200, 282-3; RFM 89, 404; BT 412; MSI 19 49).
Whether or not this holds true in general, Wittgenstein does show that
the identity of an object with itself does not provide us with an absolute
paradigm of what counts as 'doing the same' in RULE-FOLLOWING. What
counts as doing the same is determined only relative to the rule, and hence
the notion of doing the same cannot provide an independent standard: whe­
ther my saying '6' after '2, 4' counts as doing the same depends on whether
2
I follow the series y — 2x or the series y - x . There is no single, context-free
or purpose-independent way of determining what counts as doing the same.
More generally, talk about identity makes sense only relative to a practice
which lays down techniques for establishing how many things one is dealing
with, and whether or not something encountered now is the same as some­
thing encountered earlier. These criteria differ with the kind of thing at
issue. For material objects it is spatio-temporal continuity, for persons a mix­
ture of spatio-temporal continuity, memory and character-traits (PG 203;
LFM 263; BB 55, 61-2). Wittgenstein coined the term 'criterion of identity',
but the idea that talking about objects of a certain kind requires such cri­
teria goes back to Locke (An Essay concerning Human Understanding 11.16, 27)
and Frege (Foundations §§62—9). It is used in the PRIVATE LANGUAGE ARGUMENT,
168
IMAGINATION
in the Kant—Strawson attack on the notion of an immaterial substance, and
by Quine ('No entity without identity!').
imagination The British empiricists held that the only contents of the
mind are ideas or perceptions, which are understood as mental images,
and are differentiated merely by differing degrees of intensity. Kant criti­
cized this imagist conception by distinguishing between intuitions (sensa­
tions) and concepts, which are non-pictorial, and by insisting that to
possess a concept is not a matter of having a mental image, but of being
able to apply a rule. But Kant remained wedded to the imagist orthodoxy
in so far as the rules he postulated are rules for constructing mental
images. The later Wittgenstein not only denied that our mental life in gen­
eral is based on having mental images (see MEMORY; THOUGHT/THINKING;
UNDERSTANDING), but challenged the imagist conception of the imagination
itself. According to that conception, imagination is a case of non-sensory
perception, of seeing with the mind's eye: when we imagine something, we
have an image which is just like a physical picture, except that it is private
rather than public.
Like other opponents of the imagist position, such as Reid, Sartre and
Ryle, Wittgenstein did not deny the existence of mental imagery (Bilder, Vor-
stellungen), or that we can see things with the mind's eye (e.g. PI §§6, 57, II
177; RPP I §§111-19, 359, 726, 1050; RPP II §§224-39, 511; LW I §§92,
135, 315-17, 729, 794, 808; LW II 12, 19). Instead, he made four other
claims.
(a) 'One ought to ask, not what mental images (Vorstellungen) are or what
happens when one imagines something, but how the word "imagination"
(Vorstellung) is used' (PI §370). For the essence of imagination is nothing but
its GRAMMAR.
(b) Investigation of that grammar reveals that the applicability of the word
.'imagination' is not restricted to cases in which mental images cross my
mind. It is possible that people should be able to imagine things, and to
express what they imagine in writings or drawings, without seeing anything
before their mind's eye. Indeed, no mental event or process need happen
when one imagines something (RPP II §§66, 144; Z §624; see PHILOSOPHICAL
PSYCHOLOGY).
In one respect, Wittgenstein's critique of imagism does not go far enough.
Although he often uses 'to imagine' (sick vorstellen, sich denken) in a non-
perceptual way, for example when he speaks of imagining a hypothesis, an
explanation or a language (PI §§6, 19; LW I §§292, 341, 722, 777; Z §§98,
148, 440, 571), he maintains that it is essential to imagining that it employs
concepts of sense perception (RPP I §885). But it makes sense to imagine
things which it makes scant sense to perceive or picture to oneself (e.g.,
Rosa Luxemburg's last thoughts or that there are perfect numbers). It fol-
169
 IMAGINATION
lows that not all cases of imagining could involve mental imagery.
(c) Even when mental images are involved in imagination, their nature is
misunderstood by the imagist tradition. Mental images are not like physical
pictures, only private. 'A mental image (Vorstellung) is not a picture' (PI §301;
RPP H §§63, 112; Z §621; LPE 285; PR 82). It lacks definite boundaries,
and is not subject to the criteria of identity for material things. We ascribe
mental images to others on the basis of AVOWALS and BEHAVIOUR, that is, on
what the subject says or draws when asked, while in the first-person case we
employ no criterion at all. What the subject imagines, what the inner pic­
ture is of, is determined by what he says it is of, that is, there is first-person
authority in respect of mental images (PI §§377-8; LW I §811). We do not
'recognize' our mental images, nor can we observe or inspect them (PI
§§379-82; RPP II §885; Z §632). The language-game of imagining starts not
with a private entity which is then described, but with the expression of
what one imagines. A mental image is not a private entity, but the way we
imagine something, just as a visual impression is the way we organize what
we see (LW I §§440-3).
(d) Correspondingly, the imagist tradition misunderstood the relationship
between imagination and perception. 'The tie-up between imaging and
seeing is close; but there is no similarity'' (Z §625). Any description of what is
perceived can be used to describe what one imagines (although, as we have
seen, the reverse does not hold). But identical descriptions are used differ­
endy in these contexts. Perceiving and imagining are categorially distinct
(RPP II §§69-70, 130-9; Z §§629-37). (i) The difference between visual
images and visual impressions is not merely a matter of vividness, as Hume
thought (A Treatise of Human Nature I.i.3). It is unclear what standards of
vividness Hume invokes. But one may be able to imagine something more
clearly than one can see it. Furthermore, it makes no sense to wonder whe­
ther one is imagining or perceiving, although it makes sense to wonder
whether one is hallucinating or perceiving (RPP II §§96, 142; Z §621; LPP
313-14). (ii) While looking or hearing inform one about how things are in
one's environment, imagining does not (RPP II §63; Z §§621, 627). (iii)
Unlike perceiving and hallucinating, imagining is subject to the will. Thus,
one can try to 'banish' mental images in a way that one cannot visual
impressions, and one can be surprised by what one sees but not (in the
same way) by what one imagines (Z §§621, 627, 632-3). By the same token,
imagining is creative rather than receptive, and hence closer to depicting
than to seeing or hallucinating (RPP I §§111, 653; RPP II §§80-92, 115).
(iv) Wittgenstein links imagination to ASPECT-PERCEPTION: seeing X as Y often
involves imagining X as Y (PI II 193-229; RPP II §543).
induction Throughout his career Wittgenstein rejected two paradigmatic
defences of inductive reasoning against Hume's attack: the suggestion that it
170
INDUCTION
is covertly probabilistic, and the invocation of a principle of the uniformity
of nature. While the Tractatus adopts a Humean scepticism about inductive
reasoning, the later writings reject both scepticism about induction and
foundationalist attempts to vindicate it. In the Tractatus, observations on
induction (TLP 6.3f., 6.363-6.36311) are linked to the discussion of SCIENCE.
Induction is a procedure, namely of 'accepting as true the simplest law that
can be reconciled with our experiences'. It has only a psychological justifica­
tion; 'there are no grounds for believing that the simplest eventuality will in
fact be realized.' For the 'law of induction', according to which nature is
uniform - will carry on the way it has in the past - is a 'proposition with a
sense', and therefore has no logical justification. Everything outside logic, in
the domain of empirical science, is 'accidental'. In particular, CAUSATION is
neither a real nor a necessary connection between events. Consequentiy we
cannot know that the sun will rise tomorrow. For reasoning yields knowledge
only if the premises are known to be true and entail the conclusion; but the
existence of one situation never entails the existence of another. Knowledge
requires certainty, which is a limiting case of probability, that is, restricted to
tautologies. But p is a tautology, so is 'A knows thatjft' (TLP 2.012, 4.464,
5.135-5.1362, 6.36311; PT 5.04441; it would be more in fine with Wittgen­
stein's account of TAUTOLOGIES to claim that it is nonsensical, since in that
case there is nothing to be known).
Like causation, probability (TLP 5.1, 5.15-5.156; NB 8.-9.11.14; PT
5.0932) is not a real nexus between events, nor is it a special logical con­
stant peculiar to probability propositions. It is a relation between proposi­
tions, as in 'r gives to s the probability pr(r,j)'. The Tractatus provides a logical
account of probability as a relation between the structures of propositions
Which can be displayed through TRUTH-TABLES. It elaborates Laplace's classi­
cal definition of probability as a ratio of the number of possibilities which
are favourable to the occurrence of an event and the number 'of overall pos­
sibilities. Like Bolzano it drops from the definition the requirement that the
possibilities be equally likely. The degree of probability which proposition V
gives to proposition V is the ratio of the number of truth-grounds of V that
are also truth-grounds of V to the number of truth-grounds of V. The
'truth-grounds' or 'range' of a proposition are the truth-possibilities of its
arguments that make it true, those rows of its truth-table in which it has a
T. For example, 'p. q' has the truth-ground (TT), and 'p v q' has the truth-
grounds (TT), (FT), (TF):
171
INDUCTION
p 1 p.q P 1 pvq
T T T T T T
F T F F T T
T F F T F T
F F F F F F
If T is the number of T's for V, and T„ the number of T's for V where V
r
also has a T, then pr(r,j) = T „ / T . Consequentiy: (a) if V and V are logi­
r
cally incompatible, the number of truth-grounds they share is 0, and hence
pr(r,5) = 0; (b) if all the truth-grounds o f ' / are also truth-grounds of V, then
pr(r,j) = 1, that is, 's' follows from V (this is the probability that 'p.q' gives
to 'p V q'); (c) if V is either a tautology or a contradiction, then for all non-
contradictory propositions V, pr(r,s) equals 1 or 0 respectively (these are
'limiting cases of probability', TLP 5.152); (d) if neither V nor '~s' follows
from V, then 0 < pr(r,j) < 1 (depending on their inner constitutions), thus
'p v q' gives to 'p.q' the probability 1/3; (e) if V and V are logically inde­
pendent ELEMENTARY PROPOSITIONS, pr(r,j) = 1 / 2 .
Wittgenstein's account delivers the axioms of a standard a priori prob­
ability calculus. Its main problem, as for all logical theories, is to reconcile
such a priori propositions with contingent statistical observations, (e) sug­
gests, implausibly, that knowing the truth of any elementary proposition V
gives us as much reason to expect another, V, to be true as to expect V to
be false. (Only if's' is a molecular proposition can its probability relative to
the set of propositions known to be true differ from 1/2, since certain possi­
bilities may be ruled out.) Successive draws from an urn containing an equal
number of black and white balls (after each of which the ball is returned)
will show that, as the draw continues, the number of black balls drawn gra­
dually approximates to that of white balls drawn. This does not confirm the
a priori judgement that the probability of drawing a white ball is 1/2, but
rather that relative to the 'hypothesized laws of nature', and to the initial
conditions of the experiment, the two events are equipossible, that is, that a
condition for the application of the probability calculus is satisfied (TLP
5.154). This is an empirical matter, since there might be an unknown physi­
cal link between the colour of objects and their propensity to be drawn.
The probability calculus of the Tractatus collapsed with the doctrine that
there are logically independent elementary propositions, but it influenced
Ramsey, .Waismann and Carnap. The insight that there is a difference
between a priori judgements of probability and empirical statistical judge­
ments is elaborated in Philosophical Remarks (ch. XXII) and the 'Big Type­
script' (BT §§32-3; see PG 215-35; WVC 93-100; PI §§482-4). It seems that
a priori probability judgements are confirmed by statistical observations
about the relative frequency of alternative outcomes in a limited series of
172
INDUCTION
experiments. Attempts to construe inductive reasoning as a form of prob­
abilistic reasoning in which observations of past regularities render a predic­
tion probable trade on this illusion. Statistical observations, for example that
in the past 20 per cent of smokers have died of lung cancer, may lead to an
inductive extrapolation which assigns a certain probability to N.N.'s dying of
lung cancer. One cannot, however, assign a probability to the induction
itself. Further experience may confirm that the initial regularity continues,
but this merely confirms the specific extrapolation, which is itself an induc­
tive hypothesis. Consequentiy, probability cannot vindicate induction. Prob­
abilistic reasoning amounts either to statistical extrapolations (the smoker
case) which are themselves inductive, or to applications of a calculus (the
urn case) which presuppose rather than explain natural regularities and their
continuation.
Both the invocation of the law of induction and of probability try to vin­
dicate induction by assimilating it to deduction. Against this, Wittgenstein
insists that nothing in logic can license an inference from a previously
observed regularity to a universal generalization or a prediction. His later
treatment (PI §§466-90; OC passim) elaborates on the futility of trying to
vindicate induction, (a) Russell's and Ramsey's pragmatist proposal that
induction is justified through its usefulness is awry. 'Thinking has been
found to pay' itself exemplifies the pattern of reasoning it is supposed to vin­
dicate (PI §§467-9; OC §§130-1; cp. 'Limits' 148; Mathematics 245). (b) The
law of induction is merely empirical, and hence open to the same dis-
confirmation which threatens particular inductions (OC §499). (c) It plays no
role in our reasoning. We do not make a blanket assumption that what has
happened in the past must happen again. Nor does empirical reasoning fit
the deductive pattern of logical inference. We must distinguish between
'This is a fire, so it will burn me', the truth of which depends on contingent
regularities, and 'This is a fire, fire always burns, so this will burn me', the
validity of which depends not on reality but on a rule of inference (PI
§§472~3; RFM 40, 397). (d) If our empirical reasoning were deductive, this
would only relocate the problem from the empirical inference 'This is a fire,
so it will burn me' to the generalization 'Fire always burns', which rests on
exactly the same empirical grounds as the direct empirical inference (PI
§479; OC §134).
The Tractatus had concluded that inductive reasoning could not yield cer­
tainty and hence knowledge. Now Wittgenstein emphatically claims that we
do know that the sun will rise tomorrow, and can be certain that we shall
get burnt if we put our hand in the fire. This CERTAINTY is not a non-empiri­
cal umiting case of probability, nor based on an assumption like the princi­
ple of the uniformity of nature. 'It is our acting which lies at the bottom of
the language-game' of inductive reasoning (OC §§204, 273, 298, 613 19; PI
§§472—4). Our activities are informed by the collective experience of a com-
173
 INNER/OUTER
munity bound together by science and education. They are ultimately foun­
ded on our primitive reactions to the regularities of the world. One's belief
that one will be burnt is of a piece with the fear of being burnt, which is
caused by the experience of having been burnt. In the absence of causal
regularities, inductive reasoning would become irrational only in the sense
that it would lose its point, because one could never predict what was going
to happen next.
There is a naturalist streak in these observations. Unlike naturalism, how­
ever, Wittgenstein denies that our natural reactions vindicate either induc­
tion or inductive scepticism (PI §§475-83; OC §§128-9, 295-6). While
experience provides 'a hundred reasons' for our specific predictions (PI
§478), it does not provide grounds for the practice of taking relevant experi­
ences as grounds for prediction (OC §§130-1). But the demand for such
grounds is itself absurd, so that inductive scepticism gets no hold here. Witt­
genstein's attack on inductive scepticism resembles that of Strawson, but is
less clear. Inductive reasoning is not a method for predicting the future
which may be more or less adequate, it defines what it is to make rational
predictions. We call a prediction 'reasonable' precisely if it is supported by
previous experience. On a more specific level, it is a grammatical proposi­
tion that making a transition from a specific land of evidence to a certain
conclusion is rational. 'A good ground is one that looks like this' (PI §483). If
the sceptic replies that our patterns of reasoning themselves are inadequate,
because these regularities have only been observed in the past, he ignores
that there can be no such thing as now having evidence from the future
(although we can have evidence for future events) (OC §275). The sceptic's
point cannot be that there are good reasons for empirical beliefs, only past
experience is not one of them. Instead, he simply refuses to call information
about the past evidence for the future. But this could at best be a recom­
mendation for a terminological shift. Because of the AUTONOMY OF LANGUAGE,
Wittgenstein maintains that the sceptic's new rules for using the term
'reason' cannot be metaphysically superior to ours. And in pragmatic
respects they are inferior, since they remove the vital distinction between
conclusive, good and weak evidence.
inner/outer (Innen/AuSen; Inneres/Aiifieres) This contrast first appears in the
middle thirties, features prominendy in Wittgenstein's work after 1945, and
is the main theme of his last manuscripts on philosophical psychology (RPP
II §§170, 335, 643-4, 703-1; LW I & II passim). It characterizes the dualism
of the mental and the physical. We find it natural to distinguish between the
physical world containing matter, energy and tangible objects, including
human bodies, which is public, and the human mind, a private world
hidden behind our behaviour. And we think that each individual has a pri­
vileged access to his own mind, while our access to the minds of others is
174
INNER/OUTER
indirect, based on observations of their behaviour, and at best uncertain.
Wittgenstein regards this as a 'picture' which is embedded in our language.
Its roots lie in the fact that we apply mental predicates to others, but not to
ourselves, on the basis of behavioural CRITERIA, something 'external'. How­
ever, outside philosophy the distinction between the mental and the physical
does not coincide with this dichotomy of inner and outer: we regard tooth­
ache as physical pain, to be contrasted with mental suffering (LPE 278-83;
LSD 118).
The inner/outer picture informs not just Cartesian dualism, but the main­
stream of modern philosophy, including rationalism, empiricism and Kant­
ianism. Even Frege, who insisted that what we think - 'thoughts' - are
abstract entities in a 'third realm', accepted the traditional contrast between
the 'second realm' of material objects and the first realm of 'ideas' which
are the private properties of individuals: I cannot have your pain and you
cannot have my sympathy ('Thought' 68-75). Idealism and phenomenalism
dispense with the physical world, but cleave to the image of the mind as a
private immaterial theatre to which we have immediate access. BEHAVIOUR­
ISM, by contrast, reduces the mental to human behaviour, which it describes
in purely physical terms. Finally, materialism rejects the Cartesian concep­
tion of the mind as an immaterial substance, but concludes that it must be a
material substance, thereby replacing the mind/body dualism with a brain/
body dualism, in which the brain takes on the role of the inner. These posi­
tions question one half of the dichotomy, but not the contrast itself. They
ignore that we describe human behaviour not as mere bodily movement,
but ab initio in terms of our mental vocabulary, for example as jumping for
joy, chuckling with glee. The mental is neither a fiction, nor hidden behind
the outer. It infuses our behaviour and is expressed in it (LSD 10-11, 134—
5; PI §357, II 178, 222-3; LW II 24-8, 81-95).
Wittgenstein's attack on the inner/outer dichotomy is often accused of
reducing the inner to the outer, and thereby ignoring the most important
aspects of human existence. Ironically, Wittgenstein in turn accuses the
inner/outer conception of mistakenly assimilating the mental to the physical.
It construes the relationship between mental phenomena and mental terms
'on the model of material 'object and designation', and thereby turns the
mind into a realm of mental entities, states, processes and events, which are
just like their physical counterparts, only hidden and more ethereal (PI
§§293, 308, 339; B B 47, 64, 70). Like Platonism, this tendency is fuelled by
the AUGUSTINIAN PICTURE OF LANGUAGE, which suggests that all words stand for
objects, and all sentences describe something - if not physical entities, then
entities of a different kind. Wittgenstein's PRIVATE LANGUAGE ARGUMENT shows
that the idea of private entities, and hence of the mind as an inner theatre,
is incoherent. Wittgenstein also questions the assumption which unites dual­
ism, materialism arid behaviourism, namely that first-person present tense
175
 INNER/OUTER
psychological utterances are descriptions or reports - if not of a soul, then
of the brain or of behaviour. He claims that in fact they are typically AVOW­
ALS, expressions of the inner which are in some respects analogous to nat­
ural reactions, gestures, grimaces, etc.
Along with the attack on the inner realm goes a challenge to the idea
that self-knowledge is based on introspection, inward perception which is
direct and infallible. The Cartesian idea of an infallible faculty of perception
is mystifying. But it is equally awry to think of introspection as fallible, in
the vein of James (Psychology I 189-90) and contemporary anti-Cartesians.
For most mental phenomena, it does not even make sense to suppose that
their subject misperceives them or mistakes them for something else. The
possibility of a gap between its seeming to be so and its being so which
characterizes perception is absent. Moreover, there is no meaningful answer
to the question 'How do you know that you are in pain?' I do not observe,
perceive or recognize my own sensations or experiences, I simply have them
(LPE 278-80; LSD 111-12; MS 160 61). Philosophical talk of'introspection'
or an 'inner sense' is metaphorical, and once more projects onto the mental
features of the physical. There is an innocuous use of the term 'introspec­
tion'. We sometimes observe or describe our own state of mind, not in
avowing a toothache, but in special cases of self-reflection: 'I dreaded her
arrival all day long. On waking up I felt ... Then I remembered...' But in
such cases we do not operate a mysterious inner sense, we simply note more
or less skilfully, and over a period of time, how our thoughts, feelings and
moods change. In contrast to observing the outer, such observation often
alters the mental phenomena in question (PI §§585-7, LI 188, 220-1; RPP I
§§466-7; RPP LI §§156, 177, 722-8; LW I §§975-9; LPP 235).
As a result, Wittgenstein turns on its head the idea of epistemic PRIVACY,
according to which only I can know that I am in pain, while others can at
best surmise it. Because there is no such thing as misperceiving one's own
pain, or being mistaken about it, to say that I know that I am in pain is
either a nonsense, or an emphatic assertion that I am in pain. At the same
time, in the ordinary sense of 'know' others can, and often do, know that I
am in pain. It is also misleading to claim that such knowledge is 'indirect':
the sufferer does not know direcdy or indirecdy of his being in pain, he suf­
fers it, and for us there is no more direct way of knowing than by seeing
him moan and writhe. In such cases we do not infer - draw the conclusion -
that he is in pain, we see that he is suffering. Nevertheless, one might hold, I
cannot see the pain itself, only the behaviour which expresses it. But this is
like saying that I cannot see sounds or hear colours. It indicates only a cate-
gorial distinction between mental and behavioural terms, not that statements
involving the former are always inferred from those involving the latter (PI
§246; LSD 13; LW I §§767, 885).
It is tempting to protest that the mind is hidden in that there is always
176
INNER/OUTER
the possibility that others are lying or pretending. This shows that our
third-person judgements are fallible. It does not establish the sceptical con­
clusion that, in a particular case, we are or could always be mistaken.
Lying, deceit and pretence are parasitic on sincere avowals of the inner:
pretending to be in pain is not behaviour without mental accompaniment,
but behaviour plus something, a dishonest purpose, say, that makes sense
only in a language-game in which sincerity can often be taken for granted.
Nor is pretence possible in all cases, for example when someone falls into a
fire and screams with pain (LSD 10; LPE 318; PI §§249-50, LL 179-80,
229; Z §§570-1).
At the same time, the complex thoughts and feelings of some people may
be enigmatic to us, even if they do their best to reveal them. This is due to
the 'imponderability of the mental' (PI II 227-8; LW II 63-7, 87-93).
Ascriptions of subde emotions are not just defeasible, but may require close
personal acquaintance or even be undecidable. But rather than reinstate the
idea of an iron ontological curtain between the inner and the outer, these
points reinforce the link between the mind and behaviour. For the occa­
sional uncertainty of our judgements reflects an indeterminacy in our con­
cepts, which in turn is due to the complex nature of our FORM OF LIFE. That
the occasions for tiV use of some mental terms constitute a highly complex
syndrome is due to the fact that human behaviour is unpredictable, and our
reactions to it diverse and culturally relative.
The materialist version of the inner/outer dichotomy is more plausible
than the mentalist one, since it invokes not a mysterious ontological realm,
but an essential part of our body. Nevertheless, Wittgenstein's later work
challenges many versions of materialism. His insistence that mental pre­
dicates can be applied only to living organisms, in particular HUMAN BEINGS,
is incompatible with the view that it is the brain which thinks, feels or is
conscious, but also with functionalism, the view that mental states are func­
tional states of a machine. His PHILOSOPHICAL PSYCHOLOGY shows that our
mental concepts do not amount to a primitive scientific theory which could
be dispensed with in favour of something more up-to-date, as eliminative
materialism envisages.
Nevertheless, it seems plausible that mental phenomena are inner causes
of outward behaviour, and must hence be identical with neurophysiological
phenomena, that is, brain-processes or -states. However, even if one grants
this CAUSAL conception of the mind, it does not follow that psychological
statements describe neurophysiological phenomena. If Wittgenstein is right,
first-person present tense psychological utterances are by-and-large not
descriptions of anything, let alone the brain. Less controversially, what little
I know about my brain is based on fallible evidence, but that I have certain
sensations, intentions, beliefs, etc., is neither subject to error, ignorance or
doubt, nor based on evidence or observation of any kind.
177
INNER/OUTER
That one cannot substitute 'I am in pain' for 'my C-fibres are firing' in 'I
can doubt whether my C-fibres are firing' points to a category difference
between mental and neurophysiological concepts. However, 'I can doubt
whether...' constitutes an intensional context. Hence this failure of substitut-
ability is compatible with a token-token identity theory like Davidson's
anomalous monism, which claims merely that each individual mental phe­
nomenon must be identical with an individual neurophysiological event,
even though we are ignorant of what neurophysiological events correspond
to what mental ones. Nevertheless, Wittgenstein would reject this view. He
does not deny that a brain of a certain size and complexity is a precondi­
tion for the possession of mental capacities, and that some mental phenom­
ena (e.g., perceiving a flash of light) are correlated with specific
neurophysiological processes (PI §§376, 412). But he denies that there must
be a universal parallelism between the mental and the physical, even at the
level of tokens. 'No supposition seems to me more natural than that there is
no process in the brain correlated with . . . thinking; so that it would be
impossible to read off thought-processes from brain-processes.' For although
my spoken thoughts may be correlated with a series of impulses going out
from the brain, this order might proceed 'out of chaos' (Z §§608-11).
According to Wittgenstein, the mental does not supervene on the physical:
there need not be any neurophysiological difference between someone who
speaks thinkingly and someone who does not, just as there need not be any
physiological difference between seeds producing different plants. This
stance is problematic, since it amounts to denying that there must be a
causal explanation of mental processes. It may not violate any logico-meta-
physical necessities, but it is incompatible with a highly successful regulative
principle of the neuro-sciences. At the same time, Wittgenstein gives strong
reasons to suppose that many mental phenomena, notably understanding,
believing, INTENDING AND MEANING SOMETHING, are categorially distinct from
events, processes and states, since they lack the temporal properties ('genu­
ine duration', datability) of the neural states and goings-on they are sup­
posed to be identical with.
Moreover, even where neurophysiological phenomena are, as a matter of
empirical fact, correlated with mental phenomena, they are neither neces­
sary nor sufficient for the latter. Their presence does not entail that of
mental phenomena (whatever the reading of the EEG, I am not in pain
unless I feel that I am). And it is logically possible that mental phenomena
are present not just without neurophysiological accompaniments of a specific
kind (I can be in pain without any significant reading on the EEG), but
without any neurophysiological accompaniments whatsoever: 'it is imagin­
able that my skull should turn out empty when it was operated on' (OC §4;
see PI §§149-58; BB 118-20; RPP I §1063; see also UNDERSTANDING). This
specifically does not mean that we could doubt that normal human beings
178
INTENDING AND MEANING SOMETHING
have brains, since this is one of the hinge propositions which could be relin­
quished only at the price of a disintegration of our belief-system (see CER­
TAINTY). It means rather that there is no conceptual connection between
neurophysiological mechanisms and mental phenomena. Neurophysiological
concepts play no role in our explanation and application of mental terms:
third-person uses of mental terms are based on behavioural criteria, first-
person uses are not based on any criteria, let alone neurophysiological ones,
although a belief in a general connection between neurophysiological and
mental phenomena is part of our world-view (BB 47).
intending and meaning something The later Wittgenstein discusses
not only the logical mysteries surrounding iNTENnoNALrrY, but also the mental
side of intentional verbs like BELIEVING that p, intending to O and meaning a
particular object x. The last two are discussed in the final two parts of Philo­
sophical Investigations Part I (PI §§629-60 & 661-93, respectively). This is no
coincidence, since they are both linked to voluntary powers and obscured by
misconceptions of the WILL, the discussion of which occupies the preceding
sections. While intending had not received philosophical attention since Ben-
tham, meaning something was important to Wittgenstein's early work. The
Tractatus held that a propositional sign 'Fa' is projected onto a state ot affairs
in virtue of being accompanied by a proposition in the language of thought.
It is part of this METHOD OF PROJECTION that by 'a' I mean a particular object.
This is why among the subjects running through the Investigations is not only
Bedeutung, the conventional meaning of a word in a language, but also
Meinen, what a speaker means by a word on a particular occasion of utter­
ance (PI 18n, §§22, 33-5,-81, 186-8, 358, 504-13, 592).
For the later Wittgenstein, whether I intend or mean something, and if so
what, is determined neither by thought-processes nor by ostensive acts.
Intentional verbs do not signify phenomena - acts, activities, events, processes
or states, whether in the mind or in the brain. 'If God had looked into our
minds he would not have been able to see there whom we were speaking
of (PI II 217). The first argument for this astonishing claim is that inten­
tional verbs do not belong to those logical categories, since they lack 'genu­
ine duration' (see PHILOSOPHICAL PSYCHOLOGY). It may seem as if meaning
something is the act of directing one's attention towards it. But no such act
need be involved. If A simulates being in pain and says 'It will stop soon',
he can be said to mean the pain, even though there is no pain on which he
could focus his attention. Meaning something is no more a mental activity
than rising in price is an activity of butter (PI §§666-7, 693).
Wittgenstein's second argument is familiar from his treatment of UNDER­
STANDING. Mental or physical processes or states are neither necessary nor
sufficient for believing, intending or meaning something. There may be
empirical correlations between such phenomena and intentional attitudes.
179
 INTENDING AND MEANING SOMETHING
They may inform a psychologist about my 'unconscious' intentions. But they
do not determine the content of intentional attitudes, what I think, intend or
mean. Pace James and Russell {Psychology I 253-4; Analysis ch. XII), there are
no feelings or experiences which characterize intentional attitudes. Even if
one grants, as Wittgenstein does, that intentions might be accompanied by
characteristic feelings, these feelings are not the intentions they accompany
(PI §§591, 646, II 217; Z §33). As regards meaning something by what one
said, the temptation to postulate feelings is fuelled by confusing it with
meaning what one said. But not even the latter need involve feelings (e.g., of
sincerity): 'I meant what I said' does not report the result of self-observation,
but affirms one's willingness to stand by one's utterance. Normally, we
associate feelings with some expressions. But such feelings are neither neces­
sary nor sufficient for the expressions' making sense (BB 10-11; PI II 217;
RPP I §232; see ASPECT-PERCEPTION).
Intentional attitudes are not tied to mental images or words crossing one's
mind. My long-standing intentions could not consist in pertinent images or
words constantly crossing my mind. Equally, when I utter a sentence like
(1) Napoleon was impetuous
(meaning the victor of Austerlitz) a mental image might cross my mind, but
it need not. Moreover, even if I mean Napoleon I, an image of his nephew
may cross my mind, simply because the former always reminds me of the
latter. Finally, I can mean someone without knowing what he looks like. In
that case, even if an image crosses my mind, it could not possibly determine
whom I meant (PI §§663, 680, II 175-6; PG 103; BB 39-40, 142; RPP I
§§226-33). It is equally implausible to insist that I can mean Napoleon I
only if words like 'the victor of Austerlitz' cross my mind. One might claim
that to mean Napoleon I is at any rate to think of him. But while 'I
meant...' is sometimes equivalent to 'I thought of...', it is not always (PI
33n, §690, II 217). I might make an absent-minded remark about Napo­
leon I while thinking about Napoleon III. By the same token, James was
wrong to suggest that the entire thought must already be present in my
mind for me to mean-something by an utterance (PI §337; Z §1; LW I
§§843-4; Psychokgy I ch. LX).
A tempting objection is that although mental images are never necessary
or sufficient, and although no thought is necessary, verbal thoughts of a cer­
tain kind are sufficient for meaning something: if, while uttering 'Napoleon',
the words 'the loser of Sedan' cross my mind, I must mean Napoleon III.
In reply, Wittgenstein points out that (irrespective of whether it is uttered)
even the 'most explicit' expression determines what one means only if it in
turn is intended to identify the person meant (PI §641). But it need not be,
since it could be part of a recital, or a slogan crossing one's mind. More
fundamentally, mere phenomena, whether processes or states, whether
180
INTENDING AND MEANING SOMETHING
physiological or mental, are necessarily insufficient to determine one's inten­
tional attitudes - this is the kernel of truth in the Tractatus's invocation of
acts of meaning. 'It is not the picture which intends, but we must intend
something by it.' But if this intending in turn is a mere 'process', 'phenom­
enon' or 'fact', it is no less dead than the picture (Z §§236-8; PG 143-4,
148; RPP I §215). 'For no process could have the consequences of meaning'
(PI II 218). That I mean Napoleon I has the consequence that my utterance
counts as one about Napoleon I. It commits me to a certain claim, which in
turn licenses subsequent moves in the language-game. Such normative con­
sequences cannot follow from a description of my mind, brain or behaviour.
Similarly, a description of states or processes does not have the same con­
sequences as the claim that I intend to <X>. It does not characterize my $ing
as the exercise of a voluntary power for which I can be held responsible
(Wittgenstein draws the same lesson from 'Moore's paradox'; see BELIEF).
Wittgenstein's denial that intentional attitudes are states or processes of
the subject is one source of contemporary externalism, the view that what A
thinks is at least partly determined by facts 'external' and possibly unknown
to A, namely his relation to his physical (Putnam) or social (Burge) environ­
ment. Some early passages suggest that intentional attitudes are a matter of
certain phenomena in the right kind of surrounding (e.g. BB 147). But in his
mature work Wittgenstein explicidy denies that meaning something refers to
a 'family of mental and other processes' (Z §§9, 26). Instead, he mentions
three CRITERIA we use to ascribe intentional attitudes.
(a) AVOWALS. 4'S intention need not be evident from anything crossing his
mind, but is displayed by the expression of his intention, and the same
is true of his meaning something. He can be credited with intending
to <D, or with having meant Napoleon I, if this is what he avows,
either then or later, provided that there are no grounds for question­
ing his sincerity (Z §§3, 9, 53; PI §452; PG 103; BB 161; RPP I
§§579-80).
(b) Explanations. What A means is evident from how, if the occasion
arises, he explains, justifies or elaborates his utterances, what conse­
quences he derives from them, what replies and reactions he accepts
as pertinent (PG 40-5; Z §24).
(c) Context. Whom A meant by (1) may depend on the topic of the con­
versation, on 4's background, and especially on about whom he had a
reason to speak.
Unlike externalism, this account respects first-person authority, the fact
that, by and large, I cannot be mistaken about what I intend or mean. A
does not ascribe intentional attitudes to himself on the basis of these criteria,
he avows them. While others may cast doubt on the wisdom of his inten-
181
 INTENDING AND MEANING SOMETHING
tions, or the pertinence of his remark about Napoleon I, they cannot intelli­
gibly accuse him of being mistaken, but only of being dishonest (PI §§587,
679; Z §§22, 53). What matters about the context here are only facts of
which A is aware. And these contextual criteria can be defeated by sincere
avowals and explanations. Even if the conversation is about Napoleon III, A
means Napoleon I if he sincerely says so.
This is in line with Wittgenstein's functional conception of sense. The
content of an utterance does not simply depend upon the constituents and
structure of the type-sentence, it depends also on how a token of that type-
sentence is USED on a particular occasion (PI II 221). This in turn is a
matter of the 'intention' of the speaker. But these intentions are not the
bringing about of certain (perlocutionary) effects in the audience, as for
Grice, nor constituted by mental accompaniments of the utterance. By con­
trast to a recent trend, Wittgenstein does not explain linguistic meaning, or
even speaker's meaning, by reference to the intrinsic intentionality of the
mind (LW I §§17, 37). It is the ability of human agents to avow, explain
and elaborate on what they believe, intend and mean which underlies inten­
tional attitudes. We cannot mean whatever we wish by whatever signs we
use, simply by performing a mental act (PI 18n, §§508-10, 665; Z §6). How­
ever, this is not to dispute that there is such a thing as speaker's meaning.
The cases involved - meaning 'The weather is fine' by 'a b c d', or 'It's
warm here' by 'It's cold here' - are not ordinary cases of speaker's mean­
ing. They do not involve the resolution of indexicals, ambiguity or con­
textual unclarity, but are deviations from ordinary use. Even this is not
impossible. 'A word has the meaning someone has given to it' (BB 28; PI
§665). But it requires not a mental act, nor even just a simple avowal, but
an explanation of the rules according to which these words are used on this
occasion.
Wittgenstein agrees with externalism that we cannot simply avow any
intentional attitudes whatever. While the context cannot show that A is mis­
taken about what he believes, intends or means, it can render certain avow­
als and third-person ascriptions unintelligible. Contextual features do not
determine the content of intentional attitudes direcdy, as externalism says
they do, but they determine the range of intentional attitudes which can be
ascribed. One can be in agonizing pain for a split second, irrespective of the
setting, since it can be displayed in characteristic pain-behaviour. But one
cannot, for example, expect something for a split second, irrespective of the
setting, since an expectation 'is embedded in a situation from which it arises'
(PI §581). Moreover, one can intend to perform, or will, only what is, or
what one believes to be, within one's power (see PI §§614—16). Finally,
intentions are embedded in human customs and institutions. One can
intend to play chess only if the practice of playing the game exists (PI §§205,
337). For one can intend to O only if one can display or execute this inten-
182
INTENDING AND MEANING SOMETHING
tion. In the case of complex intentions this in turn presupposes a social and
historical context, since otherwise the relevant actions and utterances do not
count as expressions of the intention. Such a setting is not required for the
intentions to drink, run or go to sleep (as some externalists have it), since
these can be displayed in pre-conventional, non-linguistic behaviour. How­
ever, all but the most basic human intentions require the weave of a social
and historical FORM OF LIFE.
Whereas externalism holds that intentional attitudes are relational rather
than intrinsic, Wittgenstein holds that they are potentialities rather than
actualities. In one place he distinguishes them from states of consciousness
as 'dispositions' (Z §72; RPP II §57), which would render his position close
to that of Ryle. But although someone who intends to 0 is disposed (i.e.,
inclined) to Oing, this is not equivalent to having a disposition to <S> (RPP U
§178). For one has to find out what one's dispositions are through noting
how one is prone to react in certain circumstances. But one does not have
to find out that one intends to O. And one may have an intention (e.g., to
burst out angrily) without having a corresponding disposition, and vice
versa. More importandy, dispositions, are inevitably realized given certain
circumstances (see AWL 91). Intentional attitudes rest on abilities rather than
dispositions.
One might sum up Wittgenstein's position as follows:
A means N.N. by uttering Y at t =
x
had A been asked at t\ whom he meant, and been inclined to answer
honesdy, he would have answered 'N.N.'
But Wittgenstein rejects any analysis in terms of necessary and sufficient
conditions: 'the error lies in saying that meaning consists in something' (Z
§§16, 26, see §680; PI §§335, 678). His reasons emerge from the peculiar
way in which he approaches the subject, namely through retrospective self-
ascriptions like 'I was going to say . . . at £,' or 'When I said . . . at t I
u
meant...' One idea of his is that these do not just report a pre-existing con­
nection (between an incipient action and an intention, an utterance and its
referent), but draw such a connection (PI §§682-3; similarly PI §487 for
explanations of actions). Yet this is plausible only if the initial action or
utterance was indeterminate, for example when, without explicit delibera­
tion, I start drawing a head (Z §§8, 32).
Wittgenstein's second line is more promising. If intending/meaning some­
thing consisted in a certain fact (intrinsic or relational, actual or disposi­
tional), then remembering having intended/meant something must consist in
remembering this fact. But retrospective self-ascriptions are not based on
remembering a certain (counterfactual) fact. We don't say 'I meant/intended
... because, had you asked me, I would have said...', but the other way
183
 INTENnONALITY
around: 'Had you asked me, I would have said..., because I meant/inten­
ded ...' They do not depend on recollecting either the 'details' of the situa­
tion or 'the whole story' but are nonetheless 'semi-verifiable'. They do not
enjoy the complete authority of avowals at t\. Others may correct me by
reference to 'details of the situation'. But none of these pieces of evidence is
the remembered intention, since my memory does not rest on such evi­
dence. Others pronounce on what I intended by inference from details, or
by interpreting the situation, but I do not. If I did, I could not possess the
certainty I often have. Retrospective self-ascriptions are not reports of any
type of fact, but 'memory-reactions' (PI §§343, 636, 638; MSI 16 301).
intentionality This is the directedness of the mind towards an object,
which may or may not exist. The term was introduced by Brentano (Psycho­
kgie vow. Empirischen Standpunkt II.l.v), who maintained against the Cartesian
tradition that intentionality rather than consciousness is the distinguishing
feature of the mental. The problem of how thought and language can relate
to reality, be about something, goes back at least to Democritus. But it was
Plato who formulated the puzzle of intentionality: How can one think what
is not the case? For if it is not the case, then it does not exist, and what
does not exist is nothing. But to think nothing is not to think anything at all
(Theaetetus 189a; see PI §518).
The early Wittgenstein sought to explain how one part of reality, a pro-
positional sign consisting of mere sounds or marks, can represent another, a
fact 'outside in the world' (NB 27.10.14). Like Plato, he was exercised by the
'mystery of negation' - we can think 'how things are not' - and by the
puzzle that false thoughts represent even though what they represent is not
a fact and hence does not even exist. 'If a picture presents what-is-not-the-
case ... this only happens through it presenting that which is not the case.
For the picture says, as it were, "this is how it is not" and to the question
"How is it not?" just the positive proposition is the answer' (NB 3.11.14;
FW 24). The propositions p and ~p do not signify different entities: it is the
same fact which verifies one of them and falsifies the other (NL 94-6; NB
4.11.14; TLP 4.064). He linked this to a third puzzle, namely of how
thought can 'reach right up to reality' (TLP 2.151 If). If my thought is true,
what I think must be identical with what is the case, but if it is false, it
cannot be; yet the content of my thought is the same in both cases.
The PICTURE THEORY develops a solution to these interrelated puzzles, (a)
The possibility of falsehood and the mystery of negation. No fact need cor­
respond to the proposition as a whole, but something must correspond to
each of its elements (NAMES), namely an element (object) of the situation it
depicts (AWL 110). Moreover, it must share with that situation a LOGICAL
FORM: the names must have the same combinatorial possibilities as the
objects they stand for. Given the appropriate METHOD OF PROJECTION, the fact
184
INTENnONALITY
that the elements of the picture are related to each other in a determinate
way represents a specific configuration of objects.
(b) Reaching right up to reality. The antinomy is avoided by holding that,
whether or not my thought is true, its content is one and the same possibi­
lity, which is actualized in the first case but not in the second. What I think is
the 'sense of the proposition', the 'state of affairs' depicted, a possible combi­
nation of objects (TLP 3:11, 4.021). The possibility of that combination is
guaranteed by the proposition which 'contains' it (TLP 2.203, 3.02; NB
5.11.14), because the combinatorial possibilities of names mirror those of
objects. The world only decides whether or not the place in LOGICAL SPACE
determined by the proposition is filled.
On his return to philosophy, Wittgenstein abandoned not just logical ato­
mism, the idea that the possibility of representation rests on the existence of
sempiternal OBJECTS, but also the idea that representation presupposes an
agreement in form between a proposition and a possible state of affairs. He
continued to discuss the relationship between propositions and facts, but
now as a special case of intentionality, the 'harmony between thought and
reality' which obtains equally between beliefs, expectations, desires, etc., and
what verifies or fulfils them (PI §429; PG 142-3, 162-3; LWL 24). The
reason for this widening lies in his exposure to the causal theories of inten­
tionality of Ogden and Richards (The Meaning of Meaning) and especially Rus­
sell. According to the behaviourist account in The Analysis of Mind (chs I, III,
XII), a conscious desire is accompanied by a true belief as to its 'purpose',
that is, the state of affairs which will bring quiescence, the cessation of dis­
comfort. Such beliefs rest on inductive evidence concerning what sort of
state has in the past removed what sort of discomfort. Accordingly, a desire
is a feeling of discomfort which causally generates a 'behaviour cycle' termi­
nating in quiescence or pleasure. One knows what one desires just as one
knows what others desire, namely by inferring it from past patterns of beha­
viour. Russell also distinguished the 'objective' of a belief from 'what is
believed'. The latter, like the believing, must consist of present occurrences
in the believer, irrespective of the objective. The 'objective' of my belief that
Caesar crossed the Rubicon is a past event. However, 'what I believe'
cannot be the actual event that makes the belief true, since the latter has
long since passed. Instead, it is a related event occurring now in my mind;
otherwise, how could I say what I now believe?
Russell connects a thought and what satisfies it through a tertium quid: my
belief is verified if I recognize its objective, my desire is fulfilled if I have a
feeling of satisfaction. As Wittgenstein pointed out, this implies that 'if I
wanted to eat an apple, and someone punched me in the stomach, taking
away my appetite, then it was this punch that I originally wanted.' To avoid
this absurd consequence, Wittgenstein invokes a key element of his earlier
'picture conception' (PR 64; see LWL 9; FW 97; TLP 4.014, 4.023, 4.03).
185
 INTENTIONALITY
The relation between a thought and what satisfies it is not causal but INTER­
NAL, that is, constitutive of the relata. My belief that p could not be made
true by any fact other than that p, whatever feelings that fact may produce
in me. Equally, 'I should like an apple' does not mean 'I want whatever will
quell my feeling of nonsatisfaction and believe that an apple will do the
trick' (PI §440; PG 134).
Wittgenstein also attacks the very question of how I know what I think.
In saying that I want, say, an apple, I am stating what I want, not predicting
or conjecturing that an apple will quell a feeling of nonsatisfaction. Nor do I
read off what I want or believe from introspecting a mental state or process;
rather, I give expression to my belief or desire. Pace Russell, I cannot be mis­
taken about the content of my wish, and do not find out what it is I wish
(PI §441; BB 22). One can say 'I don't know what I want.' But typically,
these are cases not of ignorance (I have a determinate wish but don't know
what it is) but of indecision (I haven't made up my mind). However, we also
say, for example, 'I am longing for something, but don't know what', and
such locutions underlie Freud's idea of discovering our unconscious beliefs
and desires. Wittgenstein tries to defuse this locution by claiming that it is
equivalent to an 'intransitive' use of 'longing' — 'I have a feeling of longing'
- and merely displays a preference for a certain form of representation (BB
22-3, 29). But although a transitive use plus a disclaimer of knowledge may
express an objectiess mood or emotion, it often does more than this, for it
often avows ignorance, not of the intentional object, but of the cause of one's
emotional state. It is equivalent to 'I don't know what would make me con­
tent at present.' Wittgenstein righdy distinguished intentional relations from
causal ones, but ignored that what fuels psychoanalysis is not just those
cases where there is the possibility of vacuous reformulation through a mod­
ified intentional idiom, but causal and, more generally, hypothetical uses of
our existing intentional idiom. Sometimes we find out what we desire by
noticing our reactions (RPP II §3), and this provides a starting-point for
Freud's idea of unconscious desires.
During the early thirties (esp. PG chs VLT-LX) Wittgenstein provided an
account of intentionality without invoking a 'pre-established' logico-metaphy-
sical isomorphism (BT 189). The eventual upshot was Philosophical Investiga­
tions §§428-65, which is highly condensed and conceals the background of
his discussion. The basic idea is that puzzles about intentionality are meta­
physical shadows of humdrum grammatical rules. 'Like everything metaphy­
sical, the harmony between thought and reality is to be found in the
grammar of the language' (PG 162; Z §55). The belief that p is made true
by the fact that p, my wish for x is fulfilled by the event of my obtaining x,
the order to $ by O-ing; if it is false that this is red, then this is not red -
but these are not metaphysical truths about the relation between thought
and reality, they are GRAMMATICAL rules which lay down how the expression
186
INTENTIONALITY
of thoughts can be transformed into statements about what verifies, falsifies
or fulfils them (PG 162; PI §§136, 429, 458): 'The proposition that p' = 'the
proposition which the fact that p makes true'; 'The expectation that p' =
'the expectation which will be fulfilled if p'; 'It is false that p' = ^p'; 'The
order to O' = 'the order which is executed by O-ing'. Such rules also deter­
mine what is called 'the proposition that p\ 'the expectation that p\ etc., and
thereby establish connections between the concepts of proposition, expecta­
tion, etc., and those of fact, fulfilment, etc. The grammar of 'expectation' is
such that expectations are individuated by what would fulfil them, and such
that the expression of expectations involves a clause that can be converted
into a description of their fulfilment (PR 66-9; PG 150). This is not a har­
mony between a thought and a situation, but between one proposition and
another. 'It is in language that an expectation and its fulfilment make con­
tact' (PI §445; PG 140).
This point is concealed by the idea that an expectation as such is some­
thing 'unsatisfied' because it is the expectation of something 'outside the
process' of expectation (PI §438), and similarly for other types of thoughts. It
is a grammatical truth that a belief is a belief that something is the case, an
expectation an expectation that something will be the case, and that whe­
ther or not a belief is true or an expectation fulfilled is decided by how
things were, are or come to be. Nevertheless, the metaphor of satisfaction is
misleading. That my thought is 'unsatisfied' does not mean that I feel unsa­
tisfied until it is 'satisfied'; moreover, I may feel dissatisfied by the satisfac­
tion of my wish, if it is disappointing (PI §441; BB 22). One might drop the
notion of satisfaction, while insisting that thoughts are directed at something
extra-mental which fits them, just like a piston fits a cylinder (PI §439). How­
ever, the physical relation of fitting is external: a cylinder can be identified
without specifying the piston which fits it. The analogy might mean that the
same mathematical description applies to both cylinder and piston. But then
it merely restates the idea that the expression of my expectation fits the
statement of its fulfilment in that the same form of words occurs in both (PI
§429; PR 71; LWL 33; PG 134). There are not two events here, my expect­
ing him (which might be thought to consist, say, in my pacing up and
down) and his coming, which stand in a mysterious relation of fit. Rather,
'the expectation that p' and 'the expectation which will be fulfilled if p' are
two ways of referring to the same expectation. These points dissolve the
puzzle about how a thought, something I have here and now, can be about
something which is far away and in the future or past (PI §428; PG 136).
Given the order to <I>, one can say what will fulfil that order. But if we say
that it anticipates the future by ordering what will later happen, we must
add: or does not happen (PI §461), which tells us nothing about the future.
The idea that thoughts bridge a gulf in space and time is one motive
behind 'surrogationalism', the view, shared by Russell and classical empiri-
187
 INTENTIONALJTY
cism, that thought must contain a surrogate of its (often distant and non­
existent) object, a representation of what is thought. Surrogationalism is also
fuelled by a dilemma: on the one hand, thoughts are individuated by their
content; on the other, they cannot contain what is thought because the latter
is a possibly remote situation; hence, it seems, they must contain a surrogate
of it. But this violates the Tractatus's insistence that the proposition 'reaches
right up to reality'. For it implies that even when the proposition is true,
what actually obtains or transpires is not exactly the same as what one
believed, wanted or remembered, but only something similar to it, a paler
shadow (in this vein Hume - An Enquiry concerning Human Understanding II. 11 —
claimed that the vivacity of the idea I have when I remember something
never reaches that of the original impression). But there is a difference
between expecting something only similar to what actually transpires, for
example a light red patch instead of a standard red patch, and expecting
exactly what happens. And this is not a matter of comparing an actual
event with a mental duplicate: '"The report was not as loud as I had
expected" - "Then was there a louder bang in your expectation?"' (PI
§§442-3).
In the Tractatus what is thought is the state of affairs which obtains if it is
true. Still, that state of affairs is a 'shadow of the fact' which mediates
between the thought and the fact - albeit a perfect one. It has the same
form as the fact, and merely lacks the latter's existence. 'It is as if an event
even now stood in readiness before the door of reality and were then to
make its appearance in reality - like coming into a room' (PG 137; see
LWL 30; BT 104; BB 31-2, 36-7; PI §§519-21; TS302 7 8, 11-12; FW
57). In sum, the proposition determines a possible fact, while reality deter­
mines whether it is actualized. But in fact this boils down to the idea that if
a proposition makes sense, it is clear what would verify it, whether or not it
is true (PI §461). It does not require an intermediary. The temptation to
postulate such a shadow derives from the very idea of thought having a
'content'. The content of a thought is simply what is thought, namely that p.
Contents "are not intermediaries but logical constructions, projections of that-
clauses. Their reification confuses, for example, '4 believes that p' with '4
believes B.' In the latter case we have an object-accusative, two relata must
exist, one to believe, and one to be believed. The former, by contrast,
involves a disguised intentional-accusative. What we believe is that p. Here
'believe' does not express a genuine relation,' since it can apply without
there being two relata - that p may not be a fact. The second relatum
cannot be furnished by a possible state of affairs, since this implies that what
A believes is one thing (a fact) if/) is true, but another (a state of affairs) if p
is false.
Wittgenstein's PHILOSOPHICAL PSYCHOLOGY rejects Brentano's influential idea
that intentionality is the hallmark of the mental, on the grounds that it is
188
INTERNAL RELATIONS
absent not just from sensations and feelings, but also from objecdess moods
and emotions (a view he shares with Heidegger). His discussion of intention­
ality illustrates that philosophical clarifications are as complex and, occa­
sionally, as contrived - as the knots which they untie (Z §452). The reward
is a demystified picture which undermines the views of Plato, Russell and
the Tractatus, and challenges the contemporary view of intentionality as a
relation between a subject and a 'content', and the idea of a direction of fit
between thoughts and reality.
internal relations These are relations which could not fail to obtain,
since they are given with or (partly) constitutive of the terms (objects or
relata), such as white's being lighter than black. Equally, an internal prop­
erty is a property which a thing could not fail to possess, because it is
essential to its being the thing it is (TLP 4.122-4.1252; NM 116-17). The
internal or essential properties of an OBJECT are its LOGICAL FORM, they deter­
mine its combinatorial possibilities with other objects (TLP 2.01231,
2.0141). Wittgenstein picked up this term from Russell (Principles 221-6,
447-9; Essays 139-46; External ch. II; Logic 333-9). Russell, following
Moore, had insisted against Bradley diat there are external, non-essential
relations. Wittgenstein did not participate in this debate, but announced
that the 'vexed question' whether all relations are internal is 'settled' once
we realize that an internal relation between two situations is expressed by
an internal relation between the propositions depicting them (TLP 4.125f).
Internal relations are not genuine relations at all, since they cannot be
meaningfully expressed by a proposition. They are structural relations,
namely between propositions, or between propositions and the states of
affairs they depict (TLP 4.014). They show themselves if these propositions
are properly analysed. Thus, the internal relation between the numbers 1
and 2 shows itself in the fact that the latter succeeds the former in a formal
series. Equally, the internal relations of LOGICAL INFERENCE between proposi­
tions, for example between a proposition about a complex and propositions
about its components, are due to the truth-functional composition of the
relata, and show themselves if the latter are properly analysed (TLP 3.24,
5.131, 5.2f). This also reveals that there are external relations, since not all
propositions are truth-functionally related. Indeed, there must be external
relations, namely between different ELEMENTARY PROPOSITIONS (atomic states
of affairs), for these are logically independent.
Although Wittgenstein later abandoned the idea that all logical relations
are truth-functional, he continued to speak of logical relations as 'internal'
(PG 152-3; RFM 363-4; PI II 212). His abiding concern was to insist that
empirical and necessary propositions differ 'categorically' (M 86-7; LWL 9,
56-7; CE 443). Science ascribes properties and relations which can be ver­
ified or falsified empirically. Causal relations in particular can only be estab-
189
INTERNAL RELATIONS INTERNAL RELATIONS

lished empirically, by observation and induction (PI §169). Such properties the right one, the order was already deterrninate without it (LWL 30-6, 56-
and relations are contingent, and the corresponding statements, such as 9; PG 47; WVC 154-7). Secondly, there is no such thing as justifying or
2
'g= 9.81 m/sec ', 'Tom Tower is 25 m high' and 'Radioactivity causes doubting an internal relation. Since the relation is (partly) constitutive of
cancer', are corrigible. Even if these relations are physically necessary, the the relata, one cannot coherently deny that it obtains without ceasing to
corresponding propositions could in principle be refuted by new experiments talk about those relata. Consequentiy, a sceptic cannot meaningfully deny
or observations. By contrast, logic, mathematics and philosophy are con­ that the relation obtains. At most, he could reject the practice which treats
cerned with propositions which are bgicalfy necessary and hence a priori. In the two as internally related.
their cases, refutation is not extremely improbable, but inconceivable, as
with 'p 3 p\ '2 + 2 = 4' and 'White is lighter than black' (AWL 18; cp. TLP
5.1362, 6.123 If., 6.3ff.). The explanation is that these relations are internal:
two colours which do not stand in the relation mentioned, simply are not
white and black (RFM 75—6). It is impossible, therefore, to find out that
these relata do not stand in that relationship. Wittgenstein castigated empiri­
cism, notably Russell's account of intentionality, for assimilating internal to
external relations. Internal relations are not limiting cases of external ones.
Necessary propositions are not weU-corifirmed inductive generalizations, as
Mill and, occasionally, Russell suggested (PR 64; LWL 79-80; 'Limits').
Otherwise, the negation of a necessary proposition could be true, with the
absurd consequence that on some distant planet, exceptionally, white might
be darker than black.
Wittgenstein came to favour the terms 'grammatical relations' or 'gram­
matical connections' over 'internal relations' (M 87). GRAMMATICAL relations
are not relations which we establish by examining the relata, since we
could not identify the relata independentiy of the relations. The relata are
not really connected by a relation of 'fit', as that between a piston and a
cylinder, but 'belong' with each other (PI §§136, 437-9, 537). Like every­
thing metaphysical, internal relations are to be found in grammar. They
are creatures of our practice, since they are effected by the way we identify
things, for example by the fact that we call 144, and nothing else, the
square of 12 (Z §55; PG 160-1; LFM 73-85; RFM 88; see MATHEMATICAL
PROOF). Wittgenstein makes two interrelated points about grammatical rela­
tions. One is that they cannot be underpinned or explained by postulating
mediating links between the relata. This is directed against for example the
Tractatus's invocation of a 'sense' to mediate between a proposition and the
fact which verifies or falsifies it, Russell's invocation of feelings of satisfac­
tion to mediate between a desire and what fulfils it (see INTENTIONALITY), and
the invocation of an interpretation to mediate between a rule and its appli­
cation (see RULE-FOLLOWING). Such moves lead to a dilemma: either that link
is itself internally related to both of the relata, in which case it engenders a
regress, or it is only externally related, in which case it reduces the internal
relation to an external one. I do not need to add an interpretation in order
to understand an order: unless any such interpretation called forth by the
order is the right one, it hinders rather than aids understanding; but if it is

190 191
 K L
kinaesthesis This topic exercised Wittgenstein in his latest works on
PHILOSOPHICAL PSYCHOLOGY (PI II 185-6; LPP 17, 36, 72-92, 135, 157-8,
195, 202-20, 236, 256, 277-9, 309-23; RPP I §§382-408, 452, 698, 754-
98, 948; RPP II §§63, 147; Z §§477-83, 498, 503; LW I §§386-405; LW II
16-17), presumably because it is an exception to his claim that the INNER is
typically not an object of knowledge (see AVOWAL; PRIVACY). He conceded that
we can know the positions of our limbs: my assertion that my fingers are
intertwined in a certain way is fallible and correctible by others. What he
denied is a view held by psychologists like James (Psychology II chs XX,
XXIII) and Kohler (Gestalt 127-8), namely that this knowledge is based on
kinaesthetic sensations (he also rejected the related idea that I am informed
about the location of sensations by certain of their features). I could infer
the location of limbs from bodily sensations, for example, if a characteristic
rheumatic pain always accompanies a certain position of my limbs (LPP 78).
Yet, typically, we know the position of our limbs without recourse to sensa­
tions or any kind of observation, (a) Genuine sensations have duration and
intensity; such sensations rarely accompany kinaesthetic knowledge (RPP I
§§386, 771, 783; Z §478; LPP passim), (b) It is wrong to suppose that kinaes­
thetic knowledge must rest on something. Even if it is causally dependent
upon specific neurophysiological phenomena, these need not be sensed: we
can tell the direction of a sound because of the difference in effect on our
two ears, but that difference is not sensed (PI II 185; LPP 90). (c) The
empirical fact that a disruption of afferent nerves results in loss of both
kinaesthesis and sensations does not establish that sensations are essential to
the concept of kinaesthesis. Wittgenstein also claims that what results is not
simply an absence of sensations, but a specific sensation of deprivation (RPP
I §§406, 758).
192
language-game (Sprachspiel) From 1930 onward, Wittgenstein compared
axiomatic systems with chess. This analogy stems from the formalists, who
treated arithmetic as a game played with mathematical symbols. It was casti­
gated by Frege, who saw only two alternatives: either arithmetic is about
mere signs, or it is about what the signs stand for. Wittgenstein rejects this
dichotomy. Arithmetic is no more 'about' ink marks than chess is about
wooden pieces. But that does not mean that either numerals or chess-pieces
go proxy for anything. Rather, the 'meaning' of a mathematical sign, like
that of a chess piece, is the sum of the rules that determine its possible
'moves'. What differentiates applied mathematics and language from chess
and pure mathematics is merely their 'application', the way in which they
engage with other (linguistic and non-linguistic) activities (WVC 103-5, 124,
150-1, 163, 170; MS 166 28-9; Laws II §88; see NUMBERS).
The term 'language-game' is the result of Wittgenstein's extending, from
1932 onwards, the game analogy to language as a whole. It first occurs in
TS211 578 (see also BT 201; PG 62, often cited as the first occurrence, is
later, deriving from MSI 14). Initially, it is used interchangeably with 'calcu­
lus'. Its point is to draw attention to various similarities between language
and games, just as the calculus analogy highlighted similarities between
language and formal systems.
The starting-point of both analogies is that language is a rule-guided
activity, (a) Like a game, language has constitutive rules, namely those of
GRAMMAR. Unlike strategic rules, these do not determine what move/utter­
ance will bring success, but rather what is correct or makes sense, and
thereby define the game/language, (b) The meaning of a word is not an
object it stands for, but is determined by the rules governing its operation
(LWL 43-5, 59; AWL 3, 30, 44-8, 120, 151; PG 59; see MEANING-BODY). We
learn the meaning of words by learning how to use them, just as we learn
how to play chess, not by associating the pieces with objects, but by learning
how they can be moved (see USE), (c) A proposition is a move or operation in
the game of language; it would be meaningless without the system of which
it is a part. Its sense is its role in the unfolding linguistic activity (PI §§23,
199, 421; PG 130, 172; BB 42). As in the case of games, what moves are
193
 LANGUAGE-GAME
possible depends on the situation (position on the board) and for each move
certain responses are intelligible while others are excluded.
The game analogy gradually replaces the calculus analogy. This signifies
Wittgenstein's abandonment of the CALCULUS MODEL, according to which
these rules constitute a rigid, precise and definite order hidden behind the
modey appearance of language. As part of the same development, chess is
joined by less rigid games, such as ring-a-ring o' roses, as an object of com­
parison with language. Moreover, by turning to language-games, Wittgen­
stein switched attention from the geometry of a symbolism (whether
language or calculus) to its place in human practice. He used the term in
four different capacities, which (roughly) in turn occupy centre-stage.
Teaching practices Language-games are first explained (in BB 17) as 'ways of
using signs' which are simpler than those of everyday language, 'primitive
forms of language', with which 'a child begins to make use of words'. This
view evolves into the idea of a language-game as a 'system of communica­
tion' by which children 'learn' or are 'taught' their native language (PI §7;
BB 81). The fact that many words are taught by pointing at an object is
one reason for thinking that their meaning is the object pointed at; con­
versely, by investigating our teaching practices, Wittgenstein draws attention
to the fact that the relation between a name and its object is not monolithic
(PI §§8-18). Equally, how the meaning of words like 'pain', 'dream' or
'beautiful' is, or might be, learnt sheds light on the (often non-descriptive)
role of our propositions involving those terms (PI §244; LC 1-2).
Teaching practices are important to Wittgenstein, not because he engages
in armchair speculations about the 'hypothetical history' of our subsequent
linguistic practice (he condoned the logical possibility of innate linguistic
skills), but only in so far as they display distinctive features of the subsequent
use of the words, in particular EXPLANATIONS which continue to play a role as
standards of correctness (BB 12-14; PG 188). While language-games are pri­
mitive forms of language, they are supposed to be 'complete' (AWL 101; BB
81; PI §§2, 18), in the sense in which for Wittgenstein the rational numbers
are not just an incomplete subset of the real NUMBERS. Teaching practices,
by contrast, are fragments of our language. This may be one reason why
they recede in favour of fictional language-games.
Fictional language-games These are hypothetical or invented linguistic prac­
tices of a simple or primitive kind. Such 'clear and simple language-games'
serve as 'objects of comparison' (PI §130, see §§2-27). They are supposed to
shed light on our own more-complicated language-games by way of bring­
ing into sharp relief some of their features. Wittgenstein envisages at least
two ways in which this might happen. One is to build up our complicated
discourse with terms like 'truth', 'assertion', 'proposition', etc., out of more-
194
LANGUAGE-GAME
primitive language-games. This approach dominates the Brown Book, which
discusses a string of fictional language-games, without providing any philoso­
phical stage-setting, or weaving them into any line of argument. Mercifully,
this monolithic language-game method had receded by the time of Philo­
sophical Investigations.
Another strategy is to use language-games as part of a reductio ad absurdum
argument. This constructs language-games which correspond to the under­
standing of certain concepts underlying a certain philosophical theory, and
points out the contrast with our actual language-games and concepts. For
example, a language-game is set up in which 'knowing' and 'understanding'
are used to refer to states of consciousness with genuine duration, in order
to point out that we do not use them in this way. In the Investigations, Witt­
genstein employs language-games to unmask the Tractatus's doctrine of 'sim­
ples', and its method of analysis (PI §§48, 60-4). The most famous example
of this tactic is the language-game of the builders in Investigations §2 (to which
complications are added subsequendy): it consists of four words, 'block',
'pillar', 'slab' and 'beam'. These are called out by a builder A, while B, his
assistant, brings him the stone which he has learnt to bring at each call.
This 'primitive language' is supposed to fit the AUGUSTINIAN PICTURE OF LAN­
GUAGE, according to which all words are names of objects. Its obvious point
is that our language includes many uses of words besides those of naming or
calling for an object (PI §3).
Wittgenstein conceives of this tedious interaction, together with the teach­
ing which precedes it, as a 'complete primitive language', 'the whole lan­
guage of a tribe' (PI §§2, 6; BB 77). Even sympathetic readers have
protested that although- a language might have a limited vocabulary, the
builders' practice fails to qualify as one, because, firstly, their utterances lack
syntactic structure, secondly, they do not engage in conversation and,
thirdly, the interaction does not provide for a difference between meaningful
and nonsensical employments of the vocabulary. The first point assumes,
against Wittgenstein, that there can be no one-word sentences (see PROPOSI­
TION). The second would count equally against the idea of a group of solilo­
quists (PI §243), but does not hold water in either case: actually being used
in communication is not a precondition of language-hood; and in any case
the builders could use their vocabulary to pass stones between them. The
third point would disqualify the builders from speaking a language by Witt­
genstein's own lights. Whether it applies depends on whether B's reactions
differentiate between A's making a practical mistake, like calling out for a
stone which is unavailable, and A's making a linguistic mistake, that is,
employing the vocabulary in nonsensical ways, for example by saying 'Block,
Block!' Investigations §6 describes the teaching as a process of stimulus-
response conditioning, and hence does not provide room for this distinction.
But §7 does, in so far as 'naming' the stones is supposed to set up standards
195
 LANGUAGE-GAME
for the correct use of the terms. Wittgenstein also recognizes that for this
practice to count as even 'a rudimentary language' the builders must not
behave 'merely mechanically' and must be part of a form of life similar to
our's (Z §99; MS 165 94-6).
Linguistic activities Although he continued to claim that for the purposes of
understanding our own concepts, 'nothing' is more important than the con­
struction of fictitious ones (LW I §19), Wittgenstein makes less use of fic­
tional language-games after the Investigations. Instead, he focuses more on
actual linguistic activities, and describes them against the background of our
non-linguistic practices. In the same vein, Investigations §23 provides a list of
speech acts: giving orders, describing the appearance of an object, asking,
swearing, constructing an object from a description, etc. Elsewhere, Wittgen­
stein adds more-complex activities such as lying, telling stories, reporting
dreams, confessing a motive, and forming and testing hypotheses. He also
includes modes of discourse, such as making inductive predictions, talking
about physical objects or sense-impressions, and ascribing colours to objects
(PI §§249, 363, 630, II 180, 184, 224; Z §345). He also speaks of 'the lan­
guage-game with' (i.e., the use of) words such as 'game', 'proposition',
'language', 'thought', 'read' and 'pain' (PI §§71, 96, 156, 300).
Commentators have complained that Wittgenstein gives no criteria of
identity for language-games. But there is no fundamental difficulty here.
Wittgenstein distinguishes linguistic activities at different levels of generality.
What counts as the same activity (e.g., whether one needs to distinguish tell­
ing a story from telling a joke) depends on the level concerned, and on all
levels there will be borderline cases. More problematic is that Wittgenstein
uses his list of language-games to illustrate that there are 'countless kinds of
sentences' (PI §23). One might reasonably object that while, for example,
'Are you with me?' can be used to ask a question and to issue a reprimand,
types of sentences are distinguished by their grammatical moods only
(declarative, imperative, interrogative), and that this distinction not only cuts
across Wittgenstein's (questions occur in fairy-tales and scientific theory-
construction), but is more fundamental. Wittgenstein is on safer ground in
claiming that there is an irreducible 'multiplicity of language-games' (PI
§§23-4), ways of employing words which differ in philosophically significant
respects. The diverse functions of language cannot be reduced to description
or representation, as the Tractatus's doctrine of the GENERAL PROPOSITIONAL
FORM had it. Like 'game', 'language' is a FAMILY-RESEMBLANCE word: there is
no single defining feature which all games have in common, including the
games we play with words (PI §65).
Stressing the modey of language is one use to which Wittgenstein puts
actual language-games. Another is the claim that PHILOSOPHICAL confusion
arises from crossing language-games (see RFM 117-18), that is, from using
196
LANGUAGE-GAME
the words of one language-game according to rules of another. Yet another
is that like all games, language-games are AUTONOMOUS; they are not respon­
sible to external goals, unlike an activity like cooking (PG 184—5; Z §320).
Against this game analogy it has been objected that games are subject to
pragmatic standards: they can be improved according to whether they
entertain us, take too much time, etc. But similar points apply to language-
games like greeting or measuring. Another objection is that games, unlike
linguistic activities, are trivial. Yet some games play a more important role
in our lives than certain linguistic flourishes.
Language as a game There is a point at which the analogy does break down.
Unlike games, the fragments of our linguistic practice are interrelated (e.g.,
ordering and obeying), and form part of an overall system. Wittgenstein
describes this by comparing language to an 'ancient city' (PI §18; the ana­
logy occurs in Boltzmann, Physics 77 and Mauthner, Beitrdge I 26): its centre,
everyday language, is a maze of crooked streets, while its most recent addi­
tions, specialized idioms like those of chemistry or mathematics, constitute
suburbs with straight and uniform lines. Moreover, he also employs the term
'language-game' to signify this overall system. Thus, he speaks of 'the whole
language-game' and 'the human language-game', 'our language-game' (BB
108; OC §§554—9). Indeed, it is through this use of the term that he makes
his most important point: 'I shall also call the whole, consisting of language
and the actions into which it is woven, the "language-game"' (PI §7).
It is the way in which linguistic activities are interwoven with, and
embedded in, our non-linguistic practices, which makes them less trivial
than games. The linguistic activities of the builders are as crucial to their
lives as measurement and inductive reasoning are to ours. The language-
game with 'pain' is interwoven with ways of verifying third-person pain
ascriptions, but also with commiseration, etc. Our language-games are
embedded in our FORM OF LIFE, the overall practices of a linguistic commu­
nity. Because of this connection with practice, word-games like Scrabble
would not count as language-games for Wittgenstein. This idea comes more
and more to the fore in Wittgenstein's writings. Whereas at first words have
meaning within a proposition, and the game they are used in, he later said
that 'words have meaning only in the stream of life' (LW I §913). The tech­
niques of employing them are part of our natural history. But Wittgenstein
never identified the notion of a language-game with that of a form of life.
Language-games are 'part of, embedded in, a form of life (PI §§23-5).
There is also a growing emphasis on the idea that our language-games are
not subject to justification, but rooted in our natural reactions and activities
(RPP I §916; RPP II §453; OC §§402-3, 559).
Wittgenstein's neologism has been widely accepted, and extended ('the
language-game of science' or 'of religion'). Dummett has compared verifying
197
 LOGIC
a proposition to winning at chess (although Wittgenstein repudiated that
suggestion with respect to mathematical propositions - PG 289-95). Hin-
tikka has used games of seeking and finding to provide a semantics for
quantifiers.
logic Logic studies the structural features which distinguish valid from
invalid arguments. The watershed in its development was the complete axio-
matization of the predicate calculus in Begriffsschrift. Frege pioneered logicism,
the reduction of mathematics to logic, by seeking to demonstrate the deriva-
bility of arithmetic from purely logical concepts and principles of reasoning.
He overcame the limitations of syllogistic logic by exploiting an analogy
between concepts and mathematical functions to analyse propositions into
argument-expressions and function-names instead of subject and predicate.
Frege's system was axiomatic: all the truths of the predicate calculus can be
derived as theorems from its 'basic laws' according to rules of inference.
Frege understood the axioms not as analytic consequences of arbitrary defi­
nitions, but as self-evident truths about abstract entities like numbers, con­
cepts and relations which are certified by a 'logical source of knowledge'.
These axioms 'contain', in undeveloped form, all the theorems which can be
derived from them according to rules of inference (Notation §13; Laws II
App.; Posthumous 267-79). Frege's logicism foundered on the set-theoretic
paradox discerned by Russell, who then tried to escape it through his theory
of types. The system of Principia also uses (a somewhat different) function-
theoretic analysis, and is axiomatic. Russell was not clear about what vali­
dated these axioms. He wavered between arguing that they are inductively
validated by the truth of their deductive consequences, and holding (with
Frege) that they are self-evident truths known by logical intuition (Principia I
12, 59; Problems 81; 'Theory' 156-66). But what gives them their necessary
status, he could not explain. Another weakness was that to avoid the para­
doxes he had to rely on the axiom of infinity ('The number of objects in the
universe is infinite'), which looks empirical and cannot even be known to be
true.
Young Ludwig likened the development of function-theoretic logic to the
scientific revolution in the seventeenth century (RCL). He took over - and
transformed - important elements of Frege's and Russell's logical systems.
Moreover, he followed Russell in identifying philosophy with the logical
analysis of propositions (TLP 4.003f.). But his 'philosophy of logic' departed
radically from his predecessors'. With considerable chutzpah he includes
their work under the label 'the old logic', and castigates them for having
failed to clarify the nature of logic (TLP 4.1121, 4.126; NL 93; NM 109). At
the turn of the century, there were three accounts of logical truths. Accord­
ing to psychologistic logicians like Boole and Erdmann, they describe how
human beings (by and large) think, their basic mental operations, and are
198
LOGIC
determined by the nature of the human mind. Against this, Platonists like
Frege protested that logical truths are objective, and that this objectivity can
be secured only by assuming that their subject-matter - thoughts and their
structure - is not private ideas in the minds of individuals, but abstract enti­
ties inhabiting a 'third realm' beyond space and time. Finally, Russell held
that the propositions of logic are supremely general truths about the most
pervasive features of reality, a view reminiscent of Aristode's conception of
metaphysics as the most general science (Laws I Pref.; Principles 3-9, 106;
External 189-90; 'Theory' 97-101).
Wittgenstein eschews all three alternatives through a 'reflective turn' in
the tradition of Kant. Kant distinguished between 'formal logic', which
abstracts from the objects of knowledge, and 'transcendental logic', which
investigates preconditions of thinking about objects. The former consists of
analytic a priori truths. But there are also synthetic a priori truths in mathe­
matics, metaphysics and the a priori elements of science. They hold true of
experience (are synthetic) without being made true by experience (are a
priori), because they express necessary preconditions of the possibility of
experience. Wittgenstein picked this idea up from Schopenhauer and Hertz,
who explained the a priori elements of science by reference to structural fea­
tures of the way we represent objects. The Tractatus extends this idea to the
analytic truths of formal logic, while rejecting the idea of synthetic a priori
truths. Necessary propositions are neither statements about the way people
actually think, nor about the most pervasive features of reality, nor about a
Platonist hinterworld, but reflect the conditions for the possibility of empirical
representation. In contrast to Kant, these conditions no longer reside in a
mental machinery. Logic investigates the nature and limits of thought,
because it is in thought that we represent reality. But it does so by drawing
limits to the 'linguistic expression of thought' (TLP Pref.). These limits are
set by LOGICAL SYNTAX, the system of rules which determines whether a com­
bination of signs is meaningful. Logical syntax antecedes questions of truth
and falsity. It cannot be overturned by empirical propositions, since nothing
which contravenes it counts as a meaningful proposition. The special status
of necessary propositions is not due to the abstract nature of their alleged
referents, for there are no LOGICAL CONSTANTS or logical 'objects'. They are
not statements about objects of any kind, but reflect 'rules of symbolism'
(TLP 6.12ff.).
The nature of this link varies with the type of necessary proposition.
Mathematical equations are pseudo-propositions. They do not say anything
about the world, but equate signs which are equivalent by virtue of rules
governing reiterable operations (TLP 6.2ff.). Metaphysical propositions are
nonsensical. They either covertly violate logical syntax (traditional meta­
physics) or, like the pronouncements of the Tractatus itself, try to say what
can only be showri, the essential structures of reality, which must be mir-
199
 LOGIC
rored by the linguistic rules for depicting reality but cannot themselves be
depicted (TLP 3.324, 4.003, 4.12ff., 6.53f.). The only expressible necessity is
that of logical propositions, which are analytic, that is, TAUTOLOGIES (TLP
6. Iff., 6.126ff.). And they too say nothing about the world, since they com­
bine empirical propositions in such a way (according to rules governing
truth-functional operations) that all factual information cancels out (TLP
6.121).
From this linguistic perspective, the Tractatus criticizes Russell's statistical
view of modality, according to which a propositional function is possible if it
is 'sometimes true'. For \Bx\fx' can be false even if it is logically or physi­
cally possible for something to be / Moreover, that '(3x]fx' is capable of
being true presupposes that fa' makes sense, that is, that 'fx' is logically pos­
sible. Modal notions are crucial not just to distinguishing logical from
empirical propositions, but also to characterizing the latter as bipolar - they
can be true and can be false. However, the modal status of a proposition -
whether it is a tautology (necessary), a contradiction (impossible) or bipolar
(possible) - cannot be expressed by a meaningful bipolar proposition (it is
not a contingent feature) but shows itself in its structure (TLP 2.012ff., 4.464,
5.525; Introduction 165; Logic 231; see SAYING/SHOWING).
Logic, then, comprises the most general preconditions for the possibility of
symbolic, and in particular linguistic, representation - it is a 'logic of repre­
sentation' (TLP 4.015). This means that there is no such thing as a logically
defective language. Logic is a condition of sense, and there is no halfway
house between sense and nonsense. The systems of Begriffsschrift and Principia
are not in better logical order than ordinary language, as Frege and Russell
held, they are simply better at revealing that order. However, even in that
capacity they are marred by their axiomatic presentation, which evinces
what for Wittgenstein are misconceptions about the nature of logic. For this
reason, he questioned all three elements of the 'old', axiomatic logic -
axioms, theorems and logical inference.
Logical theorems need not be derived from axioms; they are vacuous
tautologies, which can be recognized as true 'from the symbol alone', by
calculating their logical properties, and hence without comparing them to
reality or deducing them from other propositions (TLP 6.113, 6.126). This
fact contains 'the whole philosophy of logic' because it also casts doubt on
the axiomatic conception of axioms and rules of inference. There are no
logical truths which are essentially 'primitive' and contain an infinite
number of essentially 'derived' theorems. All 'propositions of logic are of
equal status', namely tautological; they all say the same, namely nothing.
Indeed, the TRUTH-TABLE notation shows that, for example, '~{p.~p)\ 'p V ~p'
and 'p D p' are merely different ways of expressing one and the same tautol­
ogy ~ (TT)(/>). Similarly, the fact that logical constants like ' . ' o r 'Z>' are
interdefinable shows that they are not primitive signs. Moreover, axioms
200
LOGIC
cannot be justified by appeal to their self-evidence, since the truth of a pro­
position does not follow from the fact that it is self-evident to us (TLP
5.1363, 542f, 6.127f). Although neither tautologies nor contradictions say
anything, that a certain combination of signs is tautological or contradictory
shows something about the logical relations between propositions. For exam­
ple, that '{p.[p Z) ?)) Z) (f is a tautology (and its negation a contradiction)
shows that 'q' follows from 'p' and 'p Z) q\ and thereby provides 'the form of
a proof (TLP 6.12ff.). The rules invoked by Frege and Russell are neither
necessary for nor capable of justifying LOGICAL INFERENCE.
Wittgenstein also casts doubt on the traditional idea that logic is topic-
neutral, that the laws of thought 'are everywhere the same' (Foundations
Introd.; Posthumous 128). The Tractatus accepts that the logical operators
apply equally to all types of propositions (an assumption Wittgenstein later
questioned - see GENERALITY), but denies that the rules of inference apply
equally to empirical and non-empirical disciplines. Proof by logic deduces a
bipolar proposition from given premises. Proof in logic, on the other hand,
does not discover new truths about anything, it is merely the recognition
of yet further empty tautologies. It proves not the truth of a proposition,
but that a certain combination of signs is a tautology or an equation, and
hence part of logic or mathematics (TLP 6.1263; NM 108-9). It has been
suggested that this means that logical and MATHEMATICAL PROOF cannot
exemplify the same pattern of inference as scientific reasoning. But deriv­
ing Fermat's theorem from the conjunction of the Tanayama-Weil con­
jecture with the fact that the latter implies the former, obviously
exemplifies modus ponens. If proof in logic amounts to transforming proposi­
tions through the application of truth-functional operations (TLP 6.126),
this applies equally to transforming '~(p.~p)' into '~p v p' (which are
tautologies) and to transforming '~(p-q)' into '~p V ~q' (which are not). If it
amounts to a truth-tabular calculation that a proposition of the form
' $ 3 ? is a tautology (TLP 6.1203), this is done irrespectively of whether
<X
' $ ' and P' are themselves tautologies or bipolar propositions. A tautology
must correspond to a proof in logic just as to a proof by logic. Tractatus
6.126ff. suggests that the difference lies in the fact that proof in logic is
superfluous, since the tautological nature of a proposition will be evident
in a perspicuous notation - but so will the internal relations between
empirical propositions which underlie proof by logic. What sets proof in
logic apart is not a special pattern (logic provides all patterns of proof),
but that the acceptability of the conclusion does not depend on either
reality or the truth of a set of premises.
'Logic must take care of itself (TLP 5.473f.; NB 22.8./2.9./8.9.14). This
idea attacks Russell's attempt in the theory of types to justify the rules of
logic by reference to the meanings of signs. But although there is no doc­
trine like the theory of types which could justify logic, the latter is ineffably
201
 LOGIC
grounded in reality. Logic rests on nothing, save the essential nature of pro­
positions - their BIPOLARITY: the fact that in certain combinations (tautologies)
this bipolarity cancels out shows the 'formal' properties which language must
share with reality in order to depict it. Logic does not presuppose any logi­
cal facts or experience of logical objects. But it does presuppose that 'names
have meaning and elementary propositions have sense', that is, that proposi­
tions are bipolar pictures; and it is linked to the MYSTICAL experience 'that
the world is', namely that there are indestructible simple OBJECTS which give
NAMES their meaning (TLP 5.552f., 6.12, 6.124, 6.13).
Wittgenstein later came to think that logic does not rest on ineffable foun­
dations (see AUTONOMY OF LANGUAGE). He also abandoned the idea that logic is
confined to tautologies or truth-functional relations. He first recognized that
statements like 'Nothing can be red and green all over' are legitimate and
logically necessary without being analytical in the sense of the Tractatus, and
even contemplated calling such propositions synthetic a priori. He allowed
for logical relations which fall short of entailment (see CRITERIA) and claimed
that AVOWALS are not subject to logical operations in the same way as
empirical descriptions. Finally, he noted vis-a-vis Moore's paradox that logic
excludes as 'inadmissible' not just contradictions like 'p.~p' but also 'It is
raining, but I do not believe it' (see BELIEF). According to Wittgenstein, there
are serious gaps in a 'pure logic' which concentrates exclusively on the rules
and relations codified by formal systems like the predicate calculus; he states
and that 'logic isn't as simple as logicians think it is' (ML 10.44; RPP I
§§488-9). The Tractatus was right in linking logic to language: logic provides
us with 'norms of representation', rules for transforming symbols, for passing
from premises to conclusions. But it ignored a host of rules which Wittgen­
stein included in his later notion of GRAMMAR. He continued to use the term
'logic', but with the proviso that it includes all the constitutive rules of our
language-games (OC §§56, 501, 628). Wittgenstein's non-formal conception
of logic challenges current formal semantics, which is itself indebted to the
Tractatus and the Vienna Circle.
Although the Tractatus's conception of logic was entwined with a dubious
metaphysics of symbolism, most of its criticisms of Frege and Russell do not
presuppose the latter. By linking logic to rules of symbolism, Wittgenstein
gave it a novel 'linguistic' orientation. The claim that logical propositions
are tautologies was adopted by the logical positivists, for whom it provided a
means of doing justice to the necessity of mathematics and logic without
lapsing into Platonism or recognizing synthetic a priori truths. But the
Vienna Circle ignored the Tractatus's idea that logical necessity derives from
metaphysical structures shared by language and reality, and held instead
that necessary propositions are true by virtue of arbitrary linguistic conven­
tions. It was only later that Wittgenstein himself adopted a (radically differ­
ent) version of conventionalism (see FORM OF REPRESENTATION).
202
LOGICAL ANALYSIS
logical analysis This is the process of identifying the components of a
proposition, thought or fact, and the way in which they are combined (its
LOGICAL FORM). It gained impetus through Frege's invention of the predicate
calculus. Frege's 'concept-script' was an instrument for the derivation of
arithmetic from purely logical concepts and principles of reasoning. But it
was also supposed to free 'thought' from the tyranny of words bemoaned
since Plato (Notation Pref; Posthumous 6-7, 253, 269-70). Ordinary language
conceals the logical relations and articulations of concepts and propositions,
is rife with ambiguity and vagueness, and contains vacuous singular terms
which lead to propositions without a truth-value like
(1) The present King of France is bald.
Russell pursued a similar logicist goal, but as part of a wider, Cartesian
project. He tried to justify our scientific views by first analysing them, and
then reformulating them in a 'logically perfect language' which would give
fewer hostages to scepticism. The theory of descriptions allowed him to
avoid Meinongian entities like the round square, and to maintain, against
Frege, that the principle of bivalence holds even for propositions like (1),
which is analysed into a conjunction that is false rather than truth-valueless,
if there is no present King of France, namely 'There is one and only one
object that is the present King of France, and that object is bald.' It also
suggests that the grammatical form of subject and predicate conceals the
logical form of propositions, and allowed Russell to pursue the project of
empiricist reductionism through logical rather than psychological analysis.
Logical atomism seeks to analyse propositions into atomic propositions
which refer to sense-data. Inspired by this programme, Rtissell identified
philosophy with the logical analysis of propositions (External ffas II—III; Mysti­
cism 108-9, 148-9).
Wittgenstein took over the idea that philosophy is logical analysis and
credited the theory of descriptions with having shown that the apparent
logical form of propositions differs from their real one. He concluded that
'distrust of grammar is the first requisite for philosophizing', since grammar
engenders philosophical confusions (NL 106; see RUL 11.13; TLP Pref.,
3.323f, 4.003f.; WAM 57). However, while ordinary language 'disguises
thought', it is not logically defective, as Frege and Russell had supposed. It
is capable of 'expressing every sense' (TLP 4.002), and must hence conform
to LOGICAL SYNTAX. What is needed is not an ideal language, capable of
expressing things ordinary language cannot express, but an ideal notation
which displays the logical structure already present in ordinary propositions.
The symbolisms of Frege and Russell go some way to providing a logically
perspicuous notation (TLP 3.325). This is why, in spite of divergent concep­
tions of analysis, there is considerable agreement in detail.
Frege arguably countenanced the possibility of alternative analyses (Corre-
203

LOGICAL ANALYSIS
spondence 10; Notation §10). By contrast, the logical atomism of Russell and
Wittgenstein is committed to the idea that there are definite constituents of
propositions. Nevertheless, Wittgenstein prevaricated on this issue. He con­
sidered both the possibility that ordinary material objects might be the basic
referents of language, and the possibility that analysis might go on for ever.
Eventually, he decided that analysis must go beyond ordinary material
objects, but also come to a definite end. Although a proposition of the form
'aRb' can initially be seen as the value of different propositional functions
(e.g., of \E,)a' or \QRb'), it has 'one and only one' complete analysis, which
is not arbitrary (cp. TLP 3.25, 3.3442; NB 17.6.15 with 3.9./8.10.14,
14.6.15).
Like Russell, Wittgenstein never worked out exacdy how ordinary propo­
sitions were to be analysed; he did not even provide examples of elementary
propositions or their elements. Nevertheless, two basic ideas are clear: (a)
according to the doctrine of the general propositional form, all meaningful
propositions are truth-functions of logically independent ELEMENTARY PROPOSI­
TIONS; (b) the latter consist of logically proper names, which cannot be fur­
ther analysed since they stand for absolutely simple OBJECTS (TLP 3.20Iff,
4.22f; NB 12.10.14).
Reasonably clear also is the direction analysis should take, and its Russel­
lian tools. The first step is to analyse into singular propositions all proposi­
tions which we would ordinarily recognize as in some sense complex. Thus,
general propositions like 'All swords have a sharp blade' yield singular pro­
positions like
(2) Excalibur has a sharp blade.
Such propositions contain complex concept-words such as 'has a sharp
blade' which, following Frege, are analysable into their characteristic 'marks'
(Merkmale), the necessary and sufficient conditions for their application (TLP
4.126; OL 28-9), for example
(2') Excalibur has a blade. Excalibur is sharp.
According to Tractatus 3.24, ordinary singular terms, including proper names
like 'Excalibur', are 'contractions' of a 'symbol for a complex' into a 'simple
symbol'. They can be treated along the lines of the theory of descriptions:
thus, 'Excalibur' is replaced by a definite description - e.g., 'King Arthur's
sword' - which in turn is paraphrased as an incomplete symbol through
quantifiers and concept-words:
(2*) There is one and only one x which is King Arthur's sword that x
has a blade. that x is sharp.
(2*) contains the vague predicate 'sharp', and that conjunct might therefore
be analysed into a disjunction of determinate possibilities, for example 'that
204
LOGICAL ANALYSIS
x cuts through armour v x cuts your hand when you touch it v . . . ' Given
that Wittgenstein's version of the theory of descriptions avoids the sign of
IDENTITY, the following line of analysis emerges:
x
Q>e = ®{ix)fx = (3*)0)((/j< = {y = x)). i'x.{Q xv Q x v ...))
l 2
= (3*)(>. ¥ x . («,* v Q x v ...)). ~(3x)(3y)(fx.fy).
2
For commentators' who claim that NAMES are mere 'dummies' which do
not distinguish one object from another, analysis comes to an end with such
formulae containing quantifiers and variables; the substitution of names for
bound variables adds nothing: 'The world can be completely described by
means of fully generalized propositions, i.e. without first correlating any
name with a particular object. Then, in order to arrive at the customary
mode of expression we need simply say, after an expression like "There is
one and only one x such that...": and this x is a' (TLP 5.526; NB 17./
19.10.14). However, even if the world can be described completely through
general propositions, there must be names, because only they go proxy for
particular objects: a general description does not entail that it is a specific
thing which is the only one to have a certain property (NB 31.5.15). More­
over, general propositions like (2*) cannot be the terminus of analysis. For:
(a) they are truth-functions of, and presuppose, elementary propositions
(TLP 4.411, 5, 6ff., 6.124); (b) the Russellian analysis preserves bivalence at
the price of ambiguity: (2*) can fail to be true in two ways: either if Arthur
had no sword, or if that sword does not have a sharp blade; (c) Arthur's
sword is a complex, something we ordinarily denote by a singular term but
which actually consists of parts into which it must be analysed (NB 7./20./
23.5./20.6.15).
At this point, one must switch to what Wisdom later called 'new level
analysis', one which takes us to things of an ontologically more basic kind.
The Tractatus intimates two lines such an analysis could take. 'Every state­
ment about complexes can be resolved into a statement about their con­
stituents and into those propositions that describe the complexes completely.'
The latter say that 'the complex exists', namely by stating that its con­
stituents are so related as to form the complex. A complex consists, for
example, of a's standing-in-the-relation-^? to b. A proposition which ascribes
l
a property to it - ^{aRb)' - comes out as <&a .<&b. aRb' (TLP 2.0201, 3.24;
NL 93, 101; NM 111; NB 5.9.14; PI §60). If ~aRb, then the analysandum
does not lack a truth-value, but is false. Thus,
(3) Excalibur is in the corner
is analysed into
(3') The blade is in the corner.the hilt is in the corner.the blade is
fixed in the hilt.
205
 LOGICAL ANALYSIS
However, this analysis confronts problems concerning propositions like (2).
Most properties of the complex, including a specific weight, shape and size,
are not properties of its parts. This problem can only be avoided either by
treating these properties as constituents of the complex, an absurd idea, with
which Wittgenstein seems to have toyed (see FACT), or by analysing all dis­
course into propositions which ascribe physical properties to ultimate physi­
cal particles.
The second line of analysis intimated by the Tractatus is phenomenalism all
propositions containing GENERALITY are analysed into conjunctions or disjunc­
tions of possibilities. Thus, an existential proposition like 'There is one and
only one sword of King Arthur' is analysed into a logical sum of elementary
propositions, 'pi V p v p ...'
2 3 The disjuncts joindy exhaust the possible
experiences which would make it the case that there is a complex like King
Arthur's sword.
It is unclear how either the physicalist or the phenomenalist line is to be
pursued. How, for example, are sensory modalities other than sight to be
accounted for (presumably, public space would have to be constructed out
of visual, auditory and tactile space)? Furthermore, in the TRUTH-TABLE nota­
tion every elementary proposition is represented as a truth-function of all
others, namely as a conjunction of itself and a tautology involving all other
elementary propositions. Consequentiy, a fully analysed ordinary proposition
is a long truth-function of elementary propositions, in which all elementary
propositions, and hence all names, occur. The fully analysed proposition is
much more complicated than the unanalysed proposition with which it is
supposed to be equivalent. In order to understand the sense of humdrum
propositions like (3), we have to know either Excalibur's ultimate material
constituents, or what phenomenal states would make it the case that Excali­
bur exists. This makes the understanding of a proposition depend on knowl­
edge of (recherche) empirical facts. It sits uneasily with the idea that there
are no surprises in logic (TLP 6.125f, 6.1261), and threatens the idea that
sense antecedes matters of fact. Unsurprisingly, therefore, the Notebooks had
wavered between holding that a fully analysed proposition contains as many
names as there are ultimate elements in the state of affairs depicted (a view
shared by Moore) and holding that it contains as many as the speaker
knows to be there (NB 12./20.10./18.12.14, 18.6.15; Principia Ethica 8).
The Tractatus sweeps these problems under the carpet. Its SAYING/SHOWING
distinction prohibits descriptions of the logical form of propositions. Accord­
ingly, logical analysis is no longer supposed to result in a doctrine, as in
'Notes on Logic', but is a critical activity, namely of showing on the one
hand that empirical propositions are meaningful, and how they represent
what they do, and on the other that metaphysical propositions are non­
sensical, since they violate logical syntax (TLP 6.53-7). Yet, the Tractatus
itself does not engage in the analysis of specific propositions (with the excep-
206
LOGICAL ANALYSIS
tion of perfunctory attempts to fit recalcitrant cases, like propositions about
BELIEFS and COLOURS, into its general framework). Instead, it provides the
general framework for this activity, by sketching the GENERAL PROPOSITIONAL
FORM, features a proposition must possess to represent reality. Thus, we can
know 'without further ado' that all propositions can be analysed into truth-
functions of elementary propositions consisting of simple names. By contrast,
it leaves to the 'application of logic', that is, the analysis of actual proposi­
tions of natural language, the task of answering questions like what objects
are simple? what propositions are elementary? are there relations with
twenty-seven terms? (TLP 5.55ff.). The possible forms of elementary proposi­
tions depend on the possible combinations of objects, which we cannot list
prior to future analysis. It is not a matter of experience, but it is a matter of
future discovery. We have the capacity to form meaningful sentences, but
are ignorant of what their real meaning is, and of how they signify (TLP
4.002, 5.5562; NL 100).
In 1929, Wittgenstein stressed even more that we need to discover the
logical form of propositions. He claimed to have found out, for example,
that elementary propositions must contain real numbers (RLF 163-6,
171; WVC 42-3). Yet, it was left to Carnap's Logische Aufbau der Welt to
pursue (unsuccessfully) this programme. Wittgenstein himself soon ques­
tioned not only the atomistic model of analysis into ultimate constituents,
but the idea itself: the quasi-transcendental theory of features that language
must possess, 'dogmatically' ignores the reality of language (WVC 182-3);
the quasi-empiricist project of applied logic ignores the difference between
chemical and logical analysis. Moreover, it is based on Moore's 'hellish idea'
that it takes analysis to find out what our humdrum propositions mean
(LWL 34-5, 90; M 114; WVC 129-30; PI §§60-4). Even if we could dis­
cover Excalibur's ultimate constituents, this would contribute to our knowl­
edge of its physical make-up, but not to our understanding of the sense of
(3).
Wittgenstein now holds that there are not only no 'surprises' but also no
'discoveries' in logic or GRAMMAR, since he rejects the idea that speakers have
tacit knowledge of a complex CALCULUS or arcane logical forms (WVC 77;
LWL 16-17; PG 114-15; PI §§126-9). A 'correct logical point of view'
(TLP 4.1213) is achieved not through a quasi-geological excavation, but
through a quasi-geographical OVERVIEW, which displays features of our lin­
guistic practice that he open to view. In so far as it is legitimate, analysis
amounts either to the description of grammar, or to the substitution of one
kind of notation by another, less misleading one (though the only example
of the latter method is in his treatment of identity) (PR 51; WVC 45-7; BT
418; PI §§90-2; TS220 §98).
Wittgenstein's early conception of analysis as discovering the underlying
structure of natural languages, shorn of its logical atomism, has become one
207
 LOGICAL CONSTANTS
of the sources of modern semantics through the mediation of Catnap's Logi­
cal Syntax of Language. His recantations have been instrumental in bringing an
end to reductive analysis in both its atomistic and its empiricist version.
Nowadays, the term 'analysis' often signifies no more than the elucidation of
conceptual connections. But most analytical philosophers insist that although
analysis may not reduce propositions, it still discovers their logical form and
content. In doing so, they should confront Wittgenstein's arguments against
the calculus model and against the assumption that what is said on a given
occasion is determined solely by the logical forms and constituents of type-
sentences (see CONTEXTUALISM).
logical constants These are symbols which indicate the logical form of
propositions. In the predicate calculus, these are the propositional con­
nectives '-', ' v ' , '.' and ' 3 ' , and the quantifiers '(*)' and '(3*)'. Russell, who
coined the term, used it more widely for all the fundamental concepts of
logic, including 'relation', 'set' and 'truth'. For Russell, logic describes the
most general aspects of reality. It catalogues the logical forms of facts by
abstracting from the material components of non-logical propositions. The
symbols which survive this process are names of 'logical indefinables' or
'logical constants'. These denote 'logical objects' with which we are acquain­
ted through a 'logical experience' or 'intuition'. Two groups can be dis­
tinguished, namely the 'logical forms' of atomic facts, and the logical objects
which allegedly correspond to propositional connectives and quantifiers. Just
as elementary propositions are names of 'atomic' complexes, so the con­
nectives and quantifiers by means of which molecular propositions are
formed name constituents of 'non-atomic' complexes (Principles xv, 8-11;
'Theory' 80, 97-101).
Frege did not speak of 'logical constants'. But like Russell he conceived of
fundamental logical notions, especially the distinction between concepts and
objects, as denoting ontological categories (Foundations Introd.; 'Function' 31),
and of propositional connectives and quantifiers as names of truth-functions.
Just as ordinary concepts map objects onto truth-values, so negation is a
concept (unary function) mapping a truth-value onto the converse truth-
value, and conditionality a binary function mapping a pair of truth-values
onto a truth-value. Finally, the quantifiers are variable-binding, variable-
indexed second-level functions which map concepts (first-level functions)
onto truth-values. Thus, '* conquered Gaul' has the value T for Caesar, so
that 'Caesar did not conquer Gaul' maps the truth-value T onto F, while
'There is an x such that x conquered Gaul' maps the first-level concept 'x
conquered Gaul' onto T. For Russell, propositions are not names of truth-
values (but of complexes), hence truth-functions map atomic propositions (or
propositional functions like 'x conquered Gaul' in the case of quantifiers)
onto molecular propositions.
208
LOGICAL CONSTANTS
Like Russell, Wittgenstein includes among the 'logical constants' not just
the connectives and quantifiers, but also the identity-sign '=' and the logical
forms of elementary propositions. The idea that 'there are N O logical con­
stants' stands at the beginning of his philosophy of logic, and it is the 'fun­
damental thought' of the Tractatus that the ' "logical constants" are not
representatives; that there can be no representatives of the logic of facts'
(TLP 4.0312, 5.4-5.47; RUL 22.6.12; NB 25.12.14). The signs of logic are
not names of logical entities, whether Russell's logical objects or Frege's gen­
uine functions. By the same token, the propositions of logic are not state­
ments about entities, they describe neither abstract features of empirical
reality nor a Platonic hinterworld, but are vacuous TAUTOLOGIES which reflect
rules for the combination of signs.
At first, Wittgenstein expressed the idea that logical constants do not
represent by saying that they are all 'copulae', a kind of cement which holds
together the material components of propositions, and which remains when
the latter have been abstracted away (RUL summer 1912-1.13). Later, he
separated the discussion of the LOGICAL FORMS of elementary propositions
from that of other logical constants. The main target of the former is Rus­
sell, of the latter Frege. Both are accused of hypostatizing referents for logi­
cal propositions because of failing to see that PROPOSITIONS are not names.
Against Frege in particular he insists that the signs 'F' and 'T' no more refer
to logical objects than do brackets, but merely indicate the truth-possibilities
of propositions (TLP 4.441; NL 107). 'xRy' signifies a relation between
objects - '/> v q' does not signify an analogous relation between propositions
(TLP 5.44, 5.46f.; NL 98-101; NM 116). Propositional connectives do not
name functions; they express 'truth-operations' by means of which we gen­
erate molecular propositions out of elementary ones. The truth-value and
the sense of the results of such operations are a function of the truth-values
and senses of their bases. But the operations do not name relations between
propositions, they express internal relations between the structures of propo­
sitions by showing what has to be done to a proposition to make another
out of it (e.g., that p v q' has to be negated to obtain '-p.-q') (TLP 5.2-
5.25, 5.3; NB 17.8./29.8./22.11.16).
Wittgenstein adduces several arguments against Frege. (a) If connectives
were function-names, their argument-expressions would have to be proper
names of objects. But the arguments of truth-functions are propositions,
which are totally distinct from names (TLP 4.441).
(b) If the True and the False were objects, then Frege's method for deter­
mining, for example, the sense of the negation-sign would break down (TLP
l
4.431). Thus, provided that p' has the same truth-value as 'q (e.g., T), '~/>'
l
would have the same sense as ~q\ since both express the thought that the
True falls under the concept of negation. But obviously '~p' and '~^' have
different senses, just as '/>' and Y do.
209
 LOGICAL CONSTANTS
(c) In contrast to genuine function-signs such as 'x is red', nothing in rea­
lity corresponds to A false proposition does not correspond to a negative
fact which includes an object called 'negation'; there is no fact that corre­
sponds to it. The only effect of '"' is to reverse the truth-value of a proposi­
tion. Although 'p' and '~p' have opposite senses, both mention the same
configuration of the same objects. They can be contrasted without a sepa­
rate sign, by reversing the truth-poles - 'T p F vs. 'F p T' (TLP 2.01,
4.0621, 6.1203).
(d) A function cannot be its own argument, one cannot substitute the
function '^ is a horse' in the argument-place of is a horse'; whereas the
result of a truth-operation can be the basis of that same operation (TLP
5.251).
(e) Since a function assigns one object to another, '—p' would have to be
about negation as an object, and hence to say something different from '/>'
(TLP 5.44). By the same token, infinitely many propositions, '—p\ ' p',
etc., would follow from a single proposition p, which is absurd (TLP 5.43).
(f) Only operations, not functions, can cancel each other, or vanish:
-~p=p (TLP 5.253f, 5.441; NB 24.1.15; PT 5.0022).
l
(g) D\ *(*)', '(3*)', etc. are interdefinable (~{3x)~fx s (x)fx,
(3x)(fx.x- a) = fa); hence they neither are 'primitive signs', as Frege's and
Russell's axiomatic systems assumed, nor denote different types of functions
(TLP 5.42, 5.441).
The last point applies to the quantifiers as well as to the connectives, and
the Tractatus adds other arguments to extend the rejection of logical con­
stants to quantification (see GENERALITY) and IDENTITY. With the exception of
(b), these criticisms apply to Russell as well as Frege. But some of them can
in turn be questioned. It has been claimed that (e) and (f) assume that '/>'
and '—p' have the same sense, which goes through on Wittgenstein's con­
ception of sense as the state of affairs depicted, but not on Frege's, since '/>'
and '—p' present the same truth-value in different ways, just as '2 + 2' and
2 propositions, which in turn is the result of applying truth-operations to bipo­
'2 ' present the same number in different ways. However, this assumption is
reasonable. For '/>' and '—p' say the same, whether or not this conforms with
Frege's notion of sense; indeed, Frege himself concedes that '/> D q' expresses
l Wittgenstein's non-referential account of logical operators was accepted
the same thought as ~(p. ~q)' ('Compound' 40-6).
Point (d) may also be questioned; it does not compare like with like.
Indeed a function cannot be its own argument, but an operation cannot be
its own basis either - one cannot multiply multiplication. Moreover, just as
an operation can be applied to its own result, some functions can take as
their arguments one of their own values, and this holds true precisely for
truth-functional connectives, which have truth-values for both arguments
and values. Similarly, some functions, for example fx) = ( — \)x, cancel out:
applied to the argument 1 it yields the value — 1, for the argument —1 it
yields the value 1. However, Wittgenstein could reply that this is due to the
210
LOGICAL CONSTANTS
fact that multiplication fits Frege's conception of a function no better than
negation. Whereas it is clear how certain activities can cancel each other, it is
unclear how this can be done by adding to a sentence the name of an
entity (a function). Wittgenstein here relies on (c), the general point that the
role of propositional connectives is not to refer, but to transform proposi­
tions.
Wittgenstein tries to curb the temptation to hypostatize referents for logi­
cal terms by removing them from his ideal notation. All truth-functional
operations are reduced to a single one, joint negation, which Wittgenstein
thought capable of generating, from elementary propositions, all meaningful
propositions. But even the sign for joint negation - 'JV(£)' - does not appear
in the ideal notation, since the latter presents all propositions through
TRUTH-TABLES, without using any propositional connectives (TLP 4.44, 5.101):
instead of ' / O ? ' simply '(YT¥T)(p,q)', instead of both '—p' and 'p' simply
'(TF)(/>)'. This idea is extended to quantified propositions by treating them
as logical sums or products; and identity is expressed not through a special
sign, but through using a unique sign for every object.
At the end of this purge, only one logical constant remains, the GENERAL
PROPOSITIONAL FORM which all propositions have in common, namely that
they are pictures which state how things stand. That logical constant is
given with the bare notion of an ELEMENTARY PROPOSITION. The logical opera­
tions add nothing since they are reducible to the operation of joint negation,
that is, to conjunction and negation. Being essentially bipolar, every proposi­
tion is connected with both truth and falsity, and hence with negation, while
the possibility of asserting more than one proposition contains the idea of
conjunction. All logical operations are already contained in an elementary
proposition, fa', since the latter is equivalent to '(3x)(fx .x — a)'. Logic is a
fall-out from the essence of representation, since LOGICAL INFERENCES and logi­
cal propositions (TAUTOLOGIES) arise out of the truth-functional complexity of
lar elementary propositions (TLP 5.441, 5.47ff.; RUL summer 1912; NB
5.11.14, 5.5.15).
first by the logical positivists, and later generally. But that acceptance was
mainly based on a general abhorrence of abstract entities, not on his specific
arguments or his vision of a constant-free notation. He himself abandoned
the latter. But he extended the insight into the non-referring role of logical
terms to other signs in his attack on the AUGUSTINIAN PICTURE OF LANGUAGE.
That attack also undermines the idea that the meaning of a word is what it
stands for, and thereby removes the need to express the insight that logical
operators do not refer by saying that they have no meaning (TLP 6.126).
Wittgenstein also questioned the claim that the predicate calculus provides
adequate explanations of ordinary terms like 'not', 'and', 'all', 'if . . . then'
211
 LOGICAL FORM
(LWL 52-3; PG 55; RFM 41-3; RPP I §§269-74; Z §677; PLP 105), a
point elaborated by Strawson. Their meaning is determined not by formal
stipulations but by the way we explain and use them in everyday life, and
they can be explained through examples or even ostensively.
logical form The logical form of a proposition is its structure as para­
phrased by formal logic for the purpose of revealing those features which
matter to the validity of arguments in which it occurs. The idea goes back
to Aristode's invention of logical formalization through the use of variables.
The term 'logical form' was introduced in the nineteenth century. But it
only gained wide currency in the wake of Frege's invention of the predicate
calculus, which replaces the idea that all propositions consist of subject and
predicate with a complex function-theoretic analysis and suggests that there
are many different types of propositions, which differ in their structure or
form. Russell was the first to draw methodological conclusions from this
idea. Philosophy is logical analysis, it studies the logical form of propositions.
Since there is a fundamental identity of structure between propositions and
the facts they represent, making an inventory of the logical forms of proposi­
tions will reveal the essential structure of reality {External ch. II, 212-13;
Mysticism 75; Logic 197, 216-17, 234, 331). Russell combined these influen­
tial ideas about the importance of logical forms with idiosyncratic views
about their nature. Their discovery proceeds through abstraction from non-
logical propositions. The expressions which survive this process are variables
and 'logical constants'. Among the latter, along with propositional con­
nectives and quantifiers, are names of 'pure' or 'logical forms'. Thus, 'Plato
loves Socrates' yields 'x^>y\ These forms are completely general facts - in
our case 'Something is somehow related to something' or 'There are dual
complexes.' Under Wittgenstein's influence, Russell came to deny that logi­
cal forms are 'entities' we can name. But he continued to treat them as
'logical objects' with which we are acquainted through a 'logical experience'
akin to our acquaintance with the taste of pineapple (Principles xv, 3-11,
106; 'Theory' 97-101, 113-14, 129-31).
Wittgenstein initially accepted that philosophy is the doctrine of logical
form. He credited Russell's theory of descriptions with showing that the real
logical form of propositions differs greatly from their apparent (school-gram­
matical) one, and concluded that 'distrust of grammar is the first requisite of
philosophizing' (NL 106; TLP 4.0031). He also retained the idea that the
structure of propositions can be revealed by abstracting from their material
components.'If we replace all 'constituents' of'Plato loves Socrates' by vari­
l
ables we reach a 'logical prototype' - x<&y' - which displays the logical
form of all those propositions that describe a dual relation (TLP 3.315-
3.317; NL 93, 104). While the GENERAL PROPOSITIONAL FORM is shared by all
propositions, types of propositions are distinguished by their logical form.
212
LOGICAL FORM
At the same time, Wittgenstein claimed that Russell had imposed incon­
sistent demands on these logical forms: they had to be both facts expressed
by propositions, that is, capable of being negated ('There are dual rela­
tions'), and objects designated by names ('the dual relation'). As Wittgenstein
trenchantly remarked, they were to combine 'the useful property of being
compounded', the hallmark of propositions and FACTS, with the 'agreeable
property' of being simple, which for logical atomism is the privilege of
OBJECTS (NL 100-101, 104, 107). Wittgenstein's general target was the idea,
shared by Russell and Frege, that logical signs are names of LOGICAL CON­
STANTS, arcane entities which provide the subject-matter of LOGIC (NL 98).
As regards logical forms, he insisted that they are not objects of any kind.
At first, he characterized them as 'copulae', the logical network or cement
which holds together the material components of propositions and remains
behind when these have been abstracted away (RUL summer 1912, 1.13).
Later, he insisted that the form of a proposition is not a separate object, but
determined by the forms of its constituents.
This idea is tied up with the PICTURE THEORY. A proposition is a picture
which models reality, truly or falsely, by virtue of the relationship between
its elements representing the relationship between the elements of the situa­
tion. Such a picture possesses two essential features, firstly a METHOD OF PRO­
JECTION connecting the elements of the model with the elements of the
situation it represents, and secondly structural features which it must share
with reality in order to depict the latter. Wittgenstein referred to this shared
structure as the 'form of a picture', or its 'logical form' (NB 20./25./
29.10.14). In the Tractatus he distinguished several notions:
(a) The 'structure' of a picture is the conventionally determined way in
which its elements must be arranged in order for it to model the way
in which the elements of the situation are related (TLP 2.032, 2.15).
By definition, this structure is possessed only by the picture.
(b) Something possesses the 'pictorial form' (Form der AbbUdung) required to
depict a particular situation if it is possible to arrange its elements in a
way that mirrors the relationship between the objects of that situation;
that is, pictorial form is the possibility of that arrangement, which
means that it must be shared by picture and situation (TLP 2.15-
2.172).
(c) 'Logical form' is what any picture, of whatever pictorial form, must
share with what it depicts (TLP 2.18ff.). The picture must have the
same logico-mathematical multiplicity as the situation (TLP 4.032-
4.0412, 5.474f.; Wittgenstein attributes this notion to Hertz, see
Mechanics §418), that is (i) it must have the same number of elements
as the situation has objects, and (ii) these elements must share the
combinatorial possibilities of the objects they stand for.
213
 LOGICAL FORM
(d) The 'representational form' (Form der Darstelking) is the external 'stand­
point' from which the picture represents its subject (TLP 2.173L), the
method of representing which differs in different media. While pictor­
ial form and logical form are what A must have in common with B in
order to picture it, representational form is what distinguishes them,
preventing A from being a mere duplicate of B.
Consider the law-court model of a traffic accident which inspired the
picture theory (NB 29.9.14). The form of that model includes the spatial
relationships between the toy pram and the toy lorry; it does not include
relationships which play no role in the conventions of depiction, like that
between their weights. The three-dimensional nature of the model is part of
its pictorial form; it guarantees that spatial relationships between the toys
can represent spatial relationships between lorry and pram. But so can two-
dimensional relationships between the elements of a drawing. Here we have
two pictures of the same state of affairs with different representational
forms, that is, in different media. Both media involve features (e.g., size and
colour of the elements) which distinguish the picture from what it depicts.
Neither the two-dimensionality of the drawing nor the three-dimensionality
of the model is part of the logical form, since logical form must be
common to all pictures of the same state of affairs, irrespective of their
representational form. This logical form would not be shared with the acci­
dent by, for example, a single, stationary ball, which lacks the logical multi­
plicity required to depict it. Equally, the spatial arrangement of notes in a
score is part of its representational form, it is not shared by the music. By
contrast, the possibility of ordering distinct elements along two parameters
is shared not just by music and score (and hence is part of the score's pic­
torial form) but by the music and any representation of it (e.g., a digital
recording), and it is hence part of the score's logical form. Whatever has
pictorial form also has logical form. While not every picture is, for exam­
ple, spatial, every picture must be a 'logical picture', that is, possess a logi­
cal form. A THOUGHT is a logical picture par excellence, its only pictorial form
is logical form, which means that it requires no particular medium of
depiction (TLP 2.181-2.19, 3).
Not just pictures and what they depict have a logical form, but also their
constituents - NAMES and objects. Indeed, the logical form of a proposition is
determined by that of its constituents (NB 1.11.14; TLP 2.0233, 6.23). Just
as the form of propositions and facts is the possibility of a certain structure,
that of their constituents is the possibility of their entering into certain com­
binations. The form of a name is what it has in common with all other
names of the same logico-syntactical category, its combinatorial possibilities,
which are represented by the variables of which those names are values.
The combinatorial possibilities of a name mirror those of the object it stands
214
LOGICAL FORM
for, and thereby show something about the structure of reality (TLP
2.012ff.; NB 16.8.16). An object has both internal (structural/formal) and
external properties. Its external properties consist in its being combined with
whatever other objects it happens to be combined with. Its internal proper­
ties consist in the possibility of its combining with some objects to form
states of affairs, and the impossibility of its combining with other objects
(TLP 2.0141, 4.123). Each of these combinatorial possibilities is a form of an
object, and essential to it. Their totality is the (logical) form of the object -
the logical equivalent of the chemical valency, which determines the com-
binability of elements. Thus, being coloured is a form of visual objects,
space and time forms of all objects (TLP 2.0251; PT 2.025If.). It is an inter­
nal property of a visual object not to have a pitch, but to have some colour
(and vice versa for a note), an external property to have, for example, the
colour red. It is an internal property of all objects, including sounds, to have
some spatio-temporal location (although presumably sounds would be loca­
ted in auditory rather than visual space), an external property to have
specific spatio-temporal coordinates.
The form of an object A determines for every other object whether or not
A can combine with it. This is why if even a single object A is given, all
objects are given - they are all part of A's form (TLP 5.524). It is also why
objects constitute the 'substance of the world', that which 'subsists' indepen-
dentiy of what is the case, this substance being 'both form and content'. It is
content since, independentiy of what the facts are, the only elements of facts
are the indestructible objects. It is form since, through their own forms, the
totality of objects determines what states of affairs (combinations of objects)
are possible. This fixed order of possibilities, which is equivalent to LOGICAL
SPACE, is common to all possible worlds, and the Tractatus calls it the 'form of
the world' (TLP 2.021-2.0271). Just as each proposition must share its logi­
cal form with the state of affairs it depicts, so language, the totality of pro­
positions, must share with what it depicts the logical form, 'the form of
reality' (TLP 2.18 - apparendy equivalent to the 'form of world'). The har­
mony between language and reality which makes representation possible is
the logico-pictorial isomorphism, the structural identity, between what repre­
sents and what is represented. According to the SAYING/SHOWING distinction,
however, the logical form shared between language and reality cannot itself
be represented in meaningful bipolar propositions: an object could not lack
its combinatorial possibilities, or a fact its logical form, without ceasing to be
that particular object or fact. Instead, that, for example, red is a colour
shows itself in the logico-syntactical behaviour of 'red' in empirical proposi­
tions (TLP 3.262, 3.326). The form of an object is not itself an object, but
rather its internal properties. It is represented not by a name, but by formal
concepts - 'function', 'NUMBER', 'colour', 'sound'. A formal concept does not
denote an arcane entity of which we could have a logical experience, it is in
215
 LOGICAL INFERENCE
effect a variable which we understand as soon as we understand the signs
which are its values (TLP 4.126-4.12721). A logico-syntactical category is
nothing but an abstraction from the distinctive role of certain signs within
empirical propositions.
Wittgenstein's later work abandons the idea that LOGICAL ANALYSIS dis­
covers the hidden logical form of language, but continued to reject the
reification of logico-linguistic forms (see MEANING-BODY).
logical inference This is the derivation of a proposition, the conclusion,
from a set of other propositions, the premises, which entail the conclusion.
One of the tasks of formal LOGIC is to investigate the rules which codify such
inferences by specifying that propositions of a certain structure entail propo­
sitions of another structure. Thus, modus ponens states that all inferences of
the form
(1) p',ifp then q; ergo q
are valid. To each rule of inference corresponds a logical truth, in our case
(2) (p.(pDq))Dq.
Frege and Russell constructed axiomatic systems in which the truths of the
first-order predicate calculus are derived as theorems from a handful of
axioms through the use of a couple of inference-rules (notably modus ponens
and a principle of substitution). Their formal systems (but not all of their
informal discussions) distinguish between 'axioms' and 'rules of inference' (a
difference whose importance had been demonstrated by Lewis Carroll):
'logical laws' or 'laws of thought' do not describe how people actually think
(pace psychologism) but how they must think to think truly, and are based
on correct descriptions of logical objects and relations (Laws I xvff., §§14—25,
47-8; Posthumous 128, 145-6; Problems 40-50).
Throughout his career, Wittgenstein questioned this picture of the 'rules
of deduction' or 'laws of inference'. The Tractatus distinguishes sharply
between logical propositions and valid inferences. The former are not pro­
positions about logical entities and relations, as Frege and Russell main­
tained, but TAUTOLOGIES. (2) says nothing, since it combines its constituent
propositions in such a way that all information cancels out. 4 fortiori, it does
not say that one proposition follows from others. However, that (2) is a tau­
tology makes clear that q follows from p and p Z) q, and thus provides the
'form of a proof - modus ponens (TLP 6.1201, 6.1221, 6.1264; NM 108-9,
114, 117). This solves a puzzle which exercised Frege, namely how laws of
logic can differ, in spite of the fact that they can be derived from each other
and seem 'almost without content' ('Compound' 50). Although (2) says the
same as, for example, '(p V p) Z) p\ namely nothing, they show something
216
LOGICAL INFERENCE
different, since the fact that the former is a tautology differs from the fact
that the latter is a tautology.
'Every proposition of logic is a modus ponens represented in signs' (TLP
6.1264): if <J> entails *P, then <I> Z) must, on analysis, turn out to be a tau­
tology. All logical propositions say the same - nothing - and hence are
equivalent. Axiomatic logic is wrong to distinguish primitive axioms and
derived theorems. Moreover, it claims to prove the truth of logical proposi­
tions by applying rules of deduction to axioms. Wittgenstein protests that
this ignores the difference- between proof by logic and proof in logic: 'a logi­
cal proof of a proposition that has sense and a proof in logic must be two
entirely different things' (TLP 6.1263; NM 109).
(1') If the stove is smoking, the chimney is out of order; the stove is
smoking; ergo the chimney is out of order
deduces the truth of an empirical conclusion from that of the premises. By
contrast,
(3) (p.(pDq))Dq = (~q.(pDq))D~P
proves not so much the truth of a proposition (a tautology cannot stricdy
speaking be true, since it says nothing) as that a certain sign-combination is
a tautology, and hence part of logic. It does so without reference to any
axioms, simply by calculating 'the logical properties of symbols' (NM 108-9;
TLP 6.126). Unfortunately, the Tractatus provides conflicting accounts of this
process. Tractatus 6.126 describes it as one of 'constructing' or 'producing' a
tautology out of others through successive applications of truth-functional
operations. This description fits the axiomatic procedure better than it does
the truth-tabular decision procedure of Tractatus 6.1203, which does not
derive one tautology from another, but calculates whether a proposition has
the truth-value T for all 'truth-combinations' of its components.
Perhaps the explanation is that the Tractatus does not reject one proof-
procedure, the axiomatic, in favour of another, but only the idea that any
proof-procedure establishes truths about logical entities rather than display­
ing rules for the use of the truth-functional operators (TLP 6.126). In any
event, at the deepest level, the Tractatus proposes to dispense with logical
proof altogether. All meaningful propositions are truth-functions of logically
independent ELEMENTARY PROPOSITIONS, and can be expressed through the
TRUTH-TABLE notation. In this notation (1) can be written down as an array
of three truth-tables. Each row of these truth-tables represents a different
'truth-possibility', a possible combination of truth-values of p and q. In the
following abbreviation, the truth-value of each proposition for the four
truth-possibilities are indicated by a quadruplet of T's and F's.
(1*) (TFTF)(p,q) [p]; (TTFTX/M?) \P D q\; ergo (TTFF)(M) [q\.
217
 LOGICAL INFERENCE
The sense of a sentence is given by its 'truth-conditions', its assignment of
truth-values for each truth-possibility of its arguments. Those possibilities
that make a sentence true are its 'truth-grounds'. That 'q' follows from '/»'
and 'p Z) q' means that all truth-possibilities which are truth-grounds of both
of the first two propositions - namely the first truth-possibility - are also
truth-grounds of the last. Nowhere does a T occur for both of the premises
and an F for the conclusion, that is, it is logically impossible that the pre­
mises should all be true and the conclusion nevertheless false - the defini­
tion of entailment (TLP 4.431, 5.101-5.1241). By contrast,
(4) (TTFF)(/,, ) M; <JT¥T)(p,q) \p Z) q]; ergo (T¥T¥)(p,q) [p]
?
is a fallacy (that of asserting the consequent), since there is a truth-possibihty
(the second one) in which both premises are true and the conclusion false.
Accordingly, all entailment arises out of the complexity (truth-functional
composition) of propositions (elementary propositions have no genuine
entailments). Russell recognized that this constitutes 'an amazing simplifica­
tion of the theory of inference' ('Introduction' xvi), but failed to appreciate
the radical implications. Entailment is an LMTERNAL RELATION between propo­
sitions. But rules of inference cannot justify such relations, they are indeed
superfluous (TLP 5.13-5.132; NL 93, 100; NM 108-9). Firstly, one cannot
justify an inference like (1') by reference to (1): the latter is a mere schema
which turns into a proposition only through substituting meaningful proposi­
tions for its sentence-letters, thus producing another inference of the same
form, which cannot justify the initial one. Secondly, (1) cannot be justified
by reference to the fact that (2) is a tautology. That (2) is a tautology and
that (1) is valid are two aspects of one and the same structural relation
between premises and conclusion; neither of them provides an independent
justification of the other (such a justification cannot be provided for internal
relations in general, since the relata cannot be identified without presuppos­
ing that the relation obtains). That (2) is a tautology, or that '/>' and 'p Z) q'
entail 'q', cannot even be meaningfully said, since these are internal proper­
ties of the propositions concerned, which show themselves in an adequate
notation that displays their structure (see SAYING/SHOWING). In such a notation
we would be able to recognize all logical properties and relations by inspect­
ing empirical propositions. Ascertaining that (2) is a tautology would be un­
necessary; we could dispense both with tautologies and with the truth-
tabular decision procedure for recognizing them (TLP 6.122, 6.1262).
However, in 1929 Wittgenstein realized that it is not always possible to
analyse molecular propositions as truth-functions of elementary ones, since
there are non-truth-functional logical relations. He retained the idea that
entailment is an internal or 'grammatical' relation between premises and
conclusion; but realized that not all such relations are captured by truth-
tabular containment. What holds generally is merely the (traditional) point
218
LOGICAL INFERENCE
that the conclusion of an inference does not add new information to the
premises (PLP 371; WVC 92). Moreover, rules of inference are neither non­
sensical nor superfluous. They are GRAMMATICAL rules, norms which license
the transformation of propositions. Every grammatical rule can be employed
as a rule of inference, and we invoke such rules constantly in explaining,
justifying and criticizing such symbolic transformations. As before, rules of
inference are distinguished from tautologies. Unlike (2), which says nothing,
' "(p. (p Z) ?)) Z) q" is a tautology' is a paradigm which states that transfor­
mations of a certain type are legitimate. Equally, the law of contradiction is
not the vacuous '~(p-~p)\ but a rule which excludes as nonsensical the logi­
cal product 'p.~p' (WVC 131; AWL 137-40; RPP I §44; RPP II §732; Z
§682).
The earlier view that such rules are superfluous seems due to the idea
that the internal relations between propositions flow from the nature of the
logical operations by means of which they are constructed (TLP 6.124),
which Wittgenstein now rejects. Acknowledging rules of inference or logical
relations between propositions is now seen to be on a par with under­
standing molecular propositions and the logical operators. Rules of inference
do not follow from the meaning of the logical operators, they are partly con­
stitutive of the latter. Modus ponens, the law of excluded middle, and the law
of contradiction are also partly constitutive of the concepts of a proposition,
and of inferring. In this sense they are laws of thought: a practice which
does not conform with them, for example one based on (4) or on Prior's
round-about inference ticket, does not count as reasoning (RFM 39-41, 89,
397-8; LFM 277-8; see AUTONOMY OF LANGUAGE).
The Tractatus's truth-tabular definition of entailment influenced model
theory, which conceives of an inference as a formal relation between the
truth-conditions of propositions. Wittgenstein later reverted to the more nat­
ural idea of inferring as something people do, while insisting that .it is not a
private mental process (see THOUGHT/THINKING) but one of transforming sym­
bols according to rules. Wittgenstein's abiding conviction that 'logic must
take care of itself (TLP 5.473) is at odds with model-theoretic attempts to
justify rules of inference. One suggestion from this quarter is that such rules
must be 'sound', that is, must not permit one to infer false conclusions from
true premises. Against this, Wittgenstein claims that there is a difference
between empirical and logical inferences (RFM 40, 397). The validity of
'The stove is smoking, so the chimney is out of order' depends on the truth-
values of empirical propositions - if the chimney is in order, we abandon
that inference. (1') is not amenable to such refutation. If the chimney is in
order, we blame not (1'), but one or both of the premises. (1') is not a state­
ment about reality, but a transformation of signs according to a norm of
representation. Finally, model theory has been invoked to justify rules of
inference by reference to the semantical definitions of the logical constants.
219
I
' LOGICAL NECESSITY
This project is incompatible with Wittgenstein's rejection of MEANING-BODIES,
and with his claim that proofs in logic establish not a true description of the
world, but the tautological nature of a combination of signs. If these claims
can be sustained, they reinstate the Aristotelian idea that logical laws cannot
be justified without circularity, since they are presupposed in all reasoning.
logical necessity see FORM OF REPRESENTATION; LOGIC
logical space (logischer Raum) The term originates in Boltzmann's general­
ized thermodynamics, which treats the independent properties of a physical
system as defining separate coordinates in a multidimensional system the
points of which constitute the 'ensemble of possible states'. The Tractatus
does not define the term 'logical space', but clearly it there refers to the
ensemble of logical possibilities. Logical space stands to 'reality', the exis­
tence and non-existence of states of affairs (TLP 2.05), as the potential to
the actual. The term conveys the idea that logical possibilities form a 'logical
scaffolding' (TLP 3.42), a systematic manifold akin to a coordinate system.
The world is the 'facts in logical space' (TLP 1.13), since the contingent
existence of states of affairs is embedded in an a priori order of possibilities.
There are several dimensions to the analogy between space and the ensem­
ble of logical possibilities.
(a) A 'place' (Ort) in logical space is determined by a 'proposition' (TLP
3.4-3.42), which here means an ELEMENTARY PROPOSITION. It is a possible state
of affairs, which corresponds to the two 'truth-possibilities' of an elementary
proposition - being true or being false (TLP 4.3ff.). For n propositions there
are 2" truth-possibilities, that is, possible combinations of truth-values. If
there are only two elementary propositions, p, q, then there are four such
truth-possibihties, ways the world can be, which are represented by the rows
of a TRUTH-TABLE.
P q
T T
F T
T F
F F
(b) Just as the existence of a point in geometrical space is guaranteed
independendy of whether it is occupied or empty, namely by its coordinates,
so a place in logical space, the possibility of a state of affairs, is guaranteed
by the existence of its component objects, independendy of whether or not
that state of affairs exists (TLP 3.4-3.411). A point in the visual field is sur­
rounded by a 'colour-space', that is, it must have some colour, a note some
220
LOGICAL SPACE
pitch, an object of touch some hardness (TLP 2.0131), and so on for all
determinables. Objects must fill some region of logical space, that is, realize
some actuality in the space of possibilities, but it is an empirical matter what
place they actually fill.
(c) This analogy extends to the idea of filling space. A place in logical
space is taken up or filled if the state of affairs exists. By the same token, a
proposition can leave to the facts a 'range' (Spielraum) to fill, namely those
parts of logical space (possible states of affairs) which are compatible with its
being true (TLP 4.463, 5.5262). Any proposition divides the whole of logical
space into those truth-possibilities which agree, and those which disagree
with it. Tautologies leave to reality 'the whole' of logical space, while con­
tradictions 'fill' the whole of logical space, since they are true or false,
respectively, whatever the facts. The range a molecular proposition leaves to
the facts is determined by its 'truth-conditions', which is a partitioning of the
set of truth-possibilities into those which make it true, its 'truth-grounds',
and those which make it false. The number of truth-grounds of a proposi­
tion provides a measure of its range, and thereby of its probability (TLP
4.463f, 5.101; see INDUCTION).
(d) Finally, just as space is the field within which material objects move,
logical space is a field of possible change, namely for the changing config­
urations (combinations) of objects in facts (TLP 2.027If.). And just as mate­
rial objects have a shape, which determines their possibility of movement,
the objects of the Tractatus have a LOGICAL FORM, the possibility or impossi­
bility of their combining with other objects in a state of affairs (TLP 2.011-
2.0141, 2.0251).
It might be thought that the places in logical space include not just all
possible (existing and non-existing) states of affairs, but also their negations,
since Tractatus 4.0641 states that the 'negating proposition determines a logi­
cal place different from that of the negated proposition'. But the negating pro­
position '~p' determines a logical place by describing it as 'lying outside' that
l
of p\ which means that it really determines a logical range (see (c)) consist­
ing of all possible states of affairs except p.
Tractatus 4.463 speaks of logical space as 'infinite'. This can be understood
as claiming that there must be infinitely many states of affairs or objects
(TLP 2.0131). It has further been held that this is necessary in order to
ensure the logical independence of elementary propositions: if object A can
combine only with a finite number n of objects, that it is not combined with
n — 1 of these objects entails that it is combined with the remaining object.
But what entails an elementary proposition p„ here is a molecular proposition
of the form '~p\.~pi n~p - \- Moreover, Wittgenstein declares that it is
an open question, perhaps to be solved by the 'application of logic', whether
there are infinitely many states of affairs and objects (TLP 4.2211, 5.55ff.),
and this is in fine with his idea that logic must not depend on contingent
221
 LOGICAL SPACE
facts. Finally, it seems that his account of the GENERAL PROPOSITIONAL FORM is
successful only if the number of elementary propositions is finite (TIP 5.32).
It is implied by (b) that among objects there are no 'bachelors', that is,
that all of them are actually combined with at least one other object. This is
taken for granted by most commentators, but can be disputed by reference
to Tractatus 2.013: 'Each thing is . . . in a space of possible states of affairs.
This space I can imagine empty, but I cannot imagine the thing without the
space.' This suggests that all places of logical space might be empty because
no state of affairs exists. However, under such circumstances there would be
no propositions (which are themselves facts), and hence no linguistic repre­
sentation. Moreover, if the spaces which surround objects are analogous to a
colour-space, each object must combine with at least one object from that
space: a point in the visual field must have some colour (TLP 2.0121,
2.0131). It would seem, therefore, that the possibility of bachelors is incom­
patible with the notion of logical space after all.
It follows that there is mutual dependence between objects and logical
space. On the one hand, objects depend on logical space, since it is essential
to them to have a location within it. On the other, objects structure logical
space, since their form determines their combinatorial possibilities. The
nature of every individual object determines the totality of states of affairs in
which it can occur, hence objects in general 'contain the possibility of all
situations' (TLP 2.012, 2.0123, 2.014). Moreover, since the form of any
object determines whether or not it can combine with any other object, if
even a single object is given, all objects, and hence the whole of logical
space, are given (TLP 2.0124, 5.524). This helps to explain Tractatus 3.42: 'A
proposition can determine only one place in logical space: nevertheless the
whole of logical space must already be given by it.' The immediate reason,
alluded to in the following parentheses, is that an elementary proposition
already contains all logical operations, since it can be expressed as a truth-
function of itself with a tautology involving all other elementary propositions
(see TLP 5.47). This means that depicting any single state of affairs involves
mentioning all possible states of affairs, that is, the whole of logical space.
The underlying reason why this is the logically adequate way of expressing
elementary propositions is that the possibility of any single state of affairs,
via the forms of its constituent objects, determines what other states of
affairs are possible.
This has the unpalatable consequence that understanding one thought
requires understanding them all. It is the myth that there is, as Wittgenstein
later put it, an 'a priori order of the world', 'the order of possibilities' shared
by world and thought (PI §97). He came to hold that what is logically possi­
ble is determined by the FORM OF REPRESENTATION we adopt, not by the
essence of immutable objects, MEANING-BODIES, which impose a certain order
on our linguistic practices. At the same time, Wittgenstein continued to use
222
LOGICAL SYNTAX
the idea of geometrical space to illustrate that a logical possibility can be
actualized or not, and that a possibility of a certain logical kind can be
actualized only by something internally related to it (e.g. PR 71, 111, 216-
18, 252-3; PI §671; RPP II §64).
The Tractatus's technical apparatus (truth-possibilities, range, etc.) influ­
enced model theory and possible world semantics through Camap's notion
of an 'L-state', and the theory of probability through Waismann and
Carnap.
logical syntax Logical syntax or 'logical grammar' (TLP 3.325) is the
system of rules for the use of signs which, according to the early Wittgen­
stein, lies hidden behind the surface of language and needs to be dis­
covered by LOGICAL ANALYSIS. Logic is traditionally thought to codify patterns
of valid inference. This connection with rules was given further impetus by
the development of axiomatic systems which distinguish between axioms
and the rules of inference that lay down what formulae can be inferred
from what other formulae. Moreover, Russell's theory of types invoked
logical rules to avoid the set-theoretic paradoxes. It introduced a distinction
between sentences which are either true or false, and sentences which are
meaningless or absurd, although they may be impeccable as regards voca­
bulary and (school-grammatical) syntax. A string of signs like 'The class of
men is a man' is not false but nonsensical, since it violates logical rules
(similarly, in Husserl's Logical Investigations 'logical grammar' comprises rules
which a combination of words must respect in order to be even in the
running for truth).
The idea that meaning or sense antecedes the contrast between truth and
falsity lies at the heart of Wittgenstein's conception of LOGIC. The 'rules of
logical syntax' are 'sign-rules' [Zeichenregeln) (TLP 3.3ff, 6.02, 6.124-6.126;
NM 109; RUL 11.13). They determine whether a combination of signs
makes sense, and fall into four groups:
intra-propositional ones for combining simple names within elementary
propositions (these are roughly the rules of the theory of types);
rules for the definitions of names of complexes, which introduce abbrevia-
tory symbols;
extra-propositional ones for combining elementary propositions by truth-
functional operators (these are linked to TAUTOLOGIES and LOGICAL INFER­
ENCE);
rules for reiterable operations which result in a 'formal series' (Formenreihe),
such as the series of natural integers.
Logical syntax cannot be refuted by experience, since nothing which contra­
venes it counts as a meaningful proposition. So-called 'necessary' proposi-
223
 LOGICAL SYNTAX
LOGICAL SYNTAX

tions are not statements about a special kind of object, but reflect the rules
for representing objects in bipolar propositions. This is why philosophical
problems, which are a priori, are to be resolved by reference not to reality,
but to these rules. Philosophical theories are typically nonsensical rather
than false, since they are based on violations or misunderstandings of logical
syntax (TLP 4.002ff.).
Ordinary language engenders such confusions, because it 'disguises
thought' (TLP 4.002): its school-grammatical surface conceals the underlying
logical structure. But it is not logically defective, as Frege and Russell had
supposed. There are no more or less logical languages. Any language, any
sign-system capable of representing reality, must conform to the rules of
logical syntax. Natural languages are capable of expressing every sense.
Consequently, their propositions are 'in perfect logical order' just as they
are; 'they are not in any way logically less correct or less exact or more confused
than propositions written down ... in Russell's symbolism or any other
"Begriffsschrift". (Only it is easier for us to gather their logical form when
they are expressed in an appropriate symbolism.)' (OL 10.5.22; TLP 5.5563;
NB 17./22.6.15). To be sure, many sentences of ordinary language appear
vague or ambiguous. But this vagueness is determinate and conceals that
they contain general propositions. Any specific employment of such sen­
tences is analysable into a disjunction of possibilities, and hence does riot
violate the principle of bivalence (TLP 3.24, 5.156; NB 7.9.14, 16.-22.6.15;
see DETERMINACY OF SENSE; GENERALITY). Equally, ordinary language allows the
formulation of nonsensical pseudo-propositions, and conceals the logical
form of propositions: quantifiers look like proper names ('nobody') or pre­
dicates ('exists'), ambiguities lead to philosophical confusions ('is' functions as
copula, sign of identity and existential quantifier), and formal concepts like
'object' look like genuine concepts. However, to guard against such decep­
tion we need, not an 'ideal language' supposedly capable of expressing
things natural languages cannot express, but an 'ideal notation' or 'sign-lan­
guage' (Zeichenspracke). Such a notation is 'governed by logical grammar - by
logical syntax' (TLP 3.325); it displays the hidden logical form of ordinary
propositions.
concerned with a 'logically perfect language' ('Introduction' vs. Mathematics
App.).
The Tractatus also disagrees with another aspect of Russell's position. The
theory of types states that certain kinds of symbols cannot be sensibly com­
bined because of their meanings, that is, because they stand for certain
kinds of entities ('logical types'). Wittgenstein protests that assertions like
' "The class of lions is a lion" is nonsensical' are themselves nonsensical,
since they refer to the meaning of a proposition in order to exclude the
latter as meaningless. Equally, there are no BIPOLAR propositions about the
logical type of a symbol, for example ' "Green" is not a proper name', since
that already presupposes that one understands the mentioned symbol. For­
tunately, a theory of symbolism need not talk about meanings, since the
type of a symbol shows itself in the use of the SIGN. This is why the Tractatus
speaks of logical syntax: the rules of logic are exclusively concerned with the
combination of signs and make no reference to meaning, that is, semantics
(TLP 3.33ff., 6.126; NM 109-10).
The idea that philosophy describes logical syntax without talking about
what signs stand for (meanings) influenced Carnap's Logical Syntax of Language,
which tried to avoid the conclusion that logical syntax is ineffable through
the distinction between 'material' and 'formal mode' (see SAYING/SHOWING).
The idea of categorial rules determining the combinatorial possibilities of
signs inspired Ryle's doctrine of category mistakes. Wittgenstein himself con­
tinued to hold that the bounds of sense are drawn by linguistic rules. But he
no longer restricted logico-linguistic rules to syntax, since he recognized that
the meaning of a word is not a MEANING-BODY. For a while he continued to
hold on to a CALCULUS MODEL, according to which the rules of natural lan­
guages are hidden behind the surface. From 1931 onwards he abandoned
that idea, and with it the term 'logical syntax' in favour of 'GRAMMAR'.

The idea is to express in an appropriate symbolism what in ordinary

language leads to endless misunderstandings . . . where ordinary language
disguises logical structure, where it allows the formation of pseudo-
propositions, where it uses one term in an infinity of different meanings,
we must replace it by a symbolism which gives a clear picture of the
logical structure, excludes pseudo-propositions, and uses its terms
unambiguously. (RLF 163)

As Ramsey pointed out, Russell was mistaken in holding that the Tractatus is

224 225

M proof. It is the method of calculation which determines the sense of a pro­
m a t h e m a t i c a l proof Platonists regard mathematical proof as a means for
discovering truths about an independendy existing mathematical world. Witt­
genstein rejects this view of mathematics as 'the natural history of mathema­
tical objects'. According to him, the mathematician is an inventor rather
than a discoverer (RFM 99, 111, 137-8; LFM 22, 63-8, 82-4, 101). This by
itself is neither as original nor as outlandish as his followers and detractors
have made it out to be. Although the Platonist view is intuitively plausible,
it has been attacked by philosophers since Aristotle and by constructivist
mathematicians since Kronecker. What sets Wittgenstein's conception of
proof apart is its link with the idea that mathematics is normative.
From the very beginning, Wittgenstein distinguished sharply between
proof by LOGIC and mathematics, and proof in logic and mathematics. Proof
by logic or mathematics, for example in engineering, derives the truth of an
empirical conclusion from the truth of empirical premises according to what
for Wittgenstein are rules for the transformation of signs. By contrast, a
proof in logic or mathematics does not so much deduce the truth of one
proposition from that of another, as establish that a certain combination of
signs is a tautology or an equation, that is, belongs to logic or mathematics
respectively. To say that a necessary proposition like '2 + 2 = 4' is true is not
to say that it conforms to a necessary fact in a Platonic realm, but to say
that it is a mathematical proposition; i.e., part of our FORM OF REPRESENTATION
(for Wittgenstein there is stricdy speaking no such thing as a false mathema­
tical proposition, since propositions like '2 + 2 = 5' are not part of our form
of representation). '2 + 2 = 4' lays down what counts as an intelligible
description of reality and functions as a rule of empirical inference (e.g., T
made two pies, and then another two, hence I made four pies overall'). By
the same token, the negation of a mathematical proposition, for example,
'2 + 2 4', corresponds to a NONSENSICAL transformation of empirical propo­
sitions (e.g., 'I made two pies, and then another two, hence I did not make
four pies overall') (TLP 6.113, 6.2321; PR 250-1; AWL 200; PG 373, 392).
Such a proposition has no role within empirical reasoning, although it has a
role within MATHEMATICS, but only as part of proofs by reductio ad absurdum. To
prove that a mathematical proposition is true is to incorporate it as a GRAM-
226
MATHEMATICAL PROOF
MATICAL proposition among the 'paradigms of language' (RFM 50, 162-4,
169; LFM passim).
During the transition period, Wittgenstein added to this normativist con­
ception the idea that the sense of a mathematical proposition is given by its
position of the form 'a x b\ and hence of a numerical identity like
'25 x 25 = 625' (WVC 79; PG 370). This is analogous to the claim that the
sense of an empirical proposition is given by its method of VERIFICATION.
However, to check a mathematical proposition by calculation or proof is not
to conduct an experiment (RFM 51-9, 65-75, 192-201, 364-6, 379-98;
LFM 36-9, 71-5, 85-109, 128-30). In the case of verifying an empirical
proposition we can be surprised by brute facts. By contrast, knowing how to
prove (or disprove) a theorem is to know that one must get a certain result,
and that a different result is simply unthinkable. A mathematical proof lays
down an INTERNAL RELATION, namely between performing a certain operation
and getting a certain result (AWL 185-91, 214, 223; RFM 221, 309-10,
363): it establishes, for example, that only an operation with the result 144
counts as (is called) squaring 12. Equally, once we understand how it can be
decided whether an angle can be trisected by compass and ruler, we know
that nothing could count as trisecting an angle by compass and ruler. By
contrast to empirical propositions, the route to a mathematical proposition
cannot be described without arriving at the destination: there is no gap
between knowing how to verify it, and knowing whether it can be verified
(LFM 64).
Wittgenstein realized that this threatens to undermine the existence of
mathematical 'problems',- that is, questions which have not yet been solved
(PR 170-5). In response, he distinguished between, on the one hand, propo­
sitions and questions for which there is an established method of proof or
calculation, that is, which are part of a 'proof system', and on the other,
those for which there is not. The former can be understood without having
the solution. Thus, the question 'What is 61 X 175?' has a clear sense, even
if no one has ever performed this multiplication, because all we have to do
is apply an established set of rules. By contrast, mathematical theorems
which we do not know how to decide (e.g., Goldbach's conjecture) lack such
sense (AWL 8, 197-8; PG 366, 377; see below).
Even if mathematical equations function as norms of representation in
empirical discourse, the question is whether Wittgenstein's account of the
logical connections between different equations can do justice to pure
mathematics, the inferential aspect of mathematics. This problem also faced
the conventionalism of the Vienna Circle, who claimed that necessary pro­
positions are themselves either rules (axioms and definitions), or propositions
the truth of which follows from these conventions. As Quine showed, this
position is flawed, since it leaves unexplained the necessity by which the
227
 MATHEMATICAL PROOF
theorems follow from the stipulated conventions. According to Dummett,
Wittgenstein developed an alternative to this moderate conventionalism, one
which does not presuppose non-conventional relationships of entailment.
This 'full-blooded conventionalism' holds that the logical necessity of any
statement is always the direct expression of a linguistic convention, to which
the previously established conventions do not commit us. It seems that con­
ventionalism either has to rely on the notion of logical consequence, a meta­
physical surplus-necessity, or distorts the deductive nature of mathematics
and the compelling force of its proofs.
This interpretation is right to point out that for Wittgenstein mathema­
tical theorems are not true by virtue of conventions, but themselves rules (see
MEANING-BODY). And he considers a community which has 'applied mathe­
matics without pure mathematics', that is, accepts mathematical propositions
as norms of representation without deriving them from others - roughly the
state of mathematics before it was axiomatized by the Greeks (RFM 232—4).
But he also recognized that in our mathematics we do not simply stipulate
the theorems. The result of a calculation is a rule and yet 'not simply stipu­
lated but produced according to rules', namely of inference (RFM 228; see
LFM 101, 166). If it were otherwise, we would not need techniques of cal­
culation or proof.
Moreover, in line with the idea of proof systems, Wittgenstein dis­
tinguished between the 'necessity of the whole system', and the 'necessity in
the system', which links axioms and their consequences (LFM 142-9, 241).
This corresponds to the difference between proofs which extend an existing
proof-system, like the introduction of multiplication in Z, the set of signed
integers, and mere-'homework', proofs and calculations which merely apply
an already established technique, for example a multiplication in N, the set
of natural numbers, that has not yet been performed (PR 187; LFM 69,
238; RFM 313). In the former cases, there is no logical necessity. Expan­
sions or changes of a system of proof are not predeterrnined by the existing
rules (RFM 268-70). For by applying the old technique in a new area we
change the concepts involved (we extend the meaning of 'multiplication' by
giving sense to ' — 2 X —3'). These are new concept-formations, which may
be subject to certain standards (e.g., of a pragmatic or aesthetic kind), but to
which there are genuine alternatives, as is shown by the debates about the
introduction of negative integers and of infinitesimals.
By contrast, when we apply an established system, the results are pre­
determined. It might seem that this simply takes us back to moderate con­
ventionalism: we have an arbitrary choice in selecting a certain system of
rules (e.g., between Euclidean and Riemannian geometry), but are logically
compelled within the system we choose (just as a traveller has a choice
between several trains, but can no longer change direction once he has
boarded a particular one). However, Wittgenstein's discussion of RULE-
228
MATHEMATICAL PROOF
FOLLOWING undermines this picture of logical compulsion. The proof does
not grab us by the throat and carry us to the conclusion once we have
granted the axioms and rules of inference. At any point, we can do and say
whatever we want (within the limits of physical laws). It is just that we
would not call, for example, '1,500 X 169 = 18' a multiplication. The logical
necessity within the system boils down to the applicability of certain expres­
sions. Someone who does not acknowledge a proof or calculation within an
established system 'has parted company with us' (RFM 60; LFM 106).
This distinction preserves the idea that in mathematics we know as much
as God does (LFM 103-4): within the system we can compute anything, and
outside the system there is no fact of the matter. However, even if Wittgen­
stein avoids the spectre of a non-conventional surplus-necessity, his position
faces several problems. For one thing, he does not stick to the distinction
between the necessity within, and the non-determined extension of, a proof
system, perhaps because of his occasional flirtations with a kind of rule-
scepticism. Thus, he claims that any expansion of an irrational number is an
extension of mathematics, in spite of the fact that there is an established
technique (RFM 266-7). Furthermore, he suggests that in a 'proof I have
won through to a decision', perhaps even in the case of proofs within a
system (RFM 163, 279, 309; LFM 109, 124-5).
Even if one rejects the idea that a proof expresses cognition of facts in a
Platonic realm, this is misleading in that one does not decide to be con­
vinced by an argument. But perhaps the decision Wittgenstein has in mind
is not whether to accept the proof (having constructed a proof, we cannot
resist the conclusion without ceasing to calculate), but whether to adopt the
conclusion as a norm of representation. Having established a theorem in
Euclidean geometry, we are still at liberty to employ or reject it for the pur­
poses of terrestrial navigation. In this case, talk of 'decision' might be com­
patible with Wittgenstein's insistence that while a proof does not 'compel' us
in the sense of a logical machine, it 'guides' or 'convinces' us, much in the
sense in which for Leibniz reason inclines but does not compel (RFM 161,
187, 238-9).
The need for persuasion lies also at the heart of Wittgenstein's claim
that proofs must be 'surveyable', that is, perspicuous. We must be able to
see the connections, since these connections are not an extrinsic means for
recording a brute fact about a Platonic realm, but an integral part of the
conclusion. Mathematical proofs which go beyond the straightforward
application of an established proof system do not discover existing connec­
tions between concepts, they establish these connections. In their case,
objectivity cannot mean that getting a particular result is a criterion for
having applied the proof system, but only that the extension of the system
is surveyable (RFM 150-1, 158-9, 166, 170-5, 187, 248-9; PG 330-1).
However, the question is whether the conviction which a proof must carry
229
 MATHEMATICAL PROOF
lies in the fact that it spells out implications of our axioms and definitions
which exist prior to any attempt to construct the proof, and whether we
may refuse to accept the proof without being irrational. As regards the first
part of the question, Wittgenstein's negative answer is based on the idea
that the result of a mathematical proof is concept-formation. It modifies
existing concepts by linking them with concepts with which they were
hitherto unconnected and providing us with new CRITERIA for the applica­
tion of its constituent terms. Thus, once we accept the proof of Pythagoras'
theorem, having a hypothenuse the square over which is identical with the
sum of those over its two cathetes becomes a necessary condition for some­
thing's being a right-angled triangle. However, if the proof modifies the
concept of a right-angled triangle, it cannot be driven by the unmodified con­
cept. It has been replied that this last point is trivial, provided that the new
criteria will always coincide with the old ones ('including an angle of 90°'):
whenever we judge a figure to be a right-angled triangle by the new cri­
teria, we should also have been justified to judge it to be a right-angled tri­
angle by the old criteria. If this is the case, the proof modifies concepts
only in the sense of unfolding the commitment to the new criteria already
implicit in the existing concept. But this can only mean that understanding
the term 'right-angled triangle' is incompatible with rejecting Pythagoras'
theorem. But prior to accepting the proof, understanding that term did not
require acceptance of any statements concerning the squares over hypothe­
nuse and cathetes. Yet Wittgenstein's own conception is also problematic:
to say that each new conceptual connection modifies the meaning of 'right-
angled triangle' is at best a stipulation, and one which stretches our con­
cepts beyond breaking-point when we are told that 'each new proof in
mathematics widens the meaning of "proof"' (AWL 10, see 116-17; PG
374; cp. RFM 440).
It has been argued that even radical extensions of mathematical systems
are bound by some kind of logical entelechy: our concept of addition is
already implicit in the concept of counting. A teacher who counts as we do
but has not yet adopted our technique of addition is already implicitly com­
mitted to, for example, '7 + 5 = 12'. If she counts seven girls and five boys,
and then thirteen children altogether, she must have made what by her own
lights is a counting mistake. Wittgenstein has been defended on the grounds
that saying that she must have made a mistake already imputes our criterion,
which results from accepting '7 + 5 = 1 2 ' . This is correct, but compatible
with the idea that we know that the teacher either commits what she would
accept as a mistake, or does not count in our sense, since she regularly takes
one pupil twice. The teacher could adopt a new norm of representation
which differs from ours, but only by making cumbersome and unreasonable
assumptions, for example that although she always gets the result 12, one
pupil constantly vanishes on counting [see AUTONOMY OF LANGUAGE).
230
MATHEMATICS
Wittgenstein does not deny that there are constraints on the acceptance
of proofs, notably those of a pragmatic or aesthetic kind (RFM 370; LFM
82). But it is fair to say that he does not illuminate their workings. This
holds in particular for those proofs which neither are simple calculations nor
extend mathematics in a substantial way, but constitute the bread and
butter of mathematical research.
A final problem for Wittgenstein is the claim that we do not understand
mathematical questions or propositions such as Goldbach's conjecture in
advance of having decided them. This conclusion is inevitable if we com­
bine the idea that in mathematics there is no gap between knowing how to
prove a proposition and knowing whether it can be proven, with the dog­
matic assertion that grasping the sense of a mathematical proposition
involves knowing how it can be proven. It is usually dismissed out of hand
by an equally dogmatic invocation of compositionalism: we understand what
the terms 'even number', 'prime' and 'sum' mean, hence we must under­
stand Goldbach's conjecture, that every even number greater than 2 is the
sum of two primes. However, Wittgenstein showed that understanding the
components of a proposition and their mode of combination is not a suffi­
cient condition for understanding it (see CONTEXTUALISM). Another objection,
originating in Waismann, is that without some understanding, we could not
even set out to look for a proof. Wittgenstein anticipated this objection, and
claimed that the creative mathematician understands the proposition which
he does not yet know how to prove only in the sense in which a composer
understands a theme he wants to integrate into his composition, namely in
that he has an inkling of the techniques to be employed (RFM 314—15,
370). What Wittgenstein ignores is the straightforward point that a mathe­
matician, unlike a toddler, understands Goldbach's conjecture in the sense
of knowing how it would operate as a norm of representation, that is, he
knows what it would be to accept it as an axiom, whether or not he knows
how to prove it as a theorem.
m a t h e m a t i c s An interest in mathematics initially led Wittgenstein from
engineering to philosophy. Almost half of his work between 1929 and 1944
is in this area; and shortly before he abandoned work on it he claimed that
his 'chief contribution' had been in 'the philosophy of mathematics'.
Whereas his discussions in the early and transition period include highly
technical details, he later concentrates on questions which can be illustrated
by reference to elementary arithmetic (LFM 13-14). Exegesis and evaluation
of his contributions (both early and late) are still at a rudimentary stage.
Wittgenstein's claims about mathematics are often baffling, and have been
accused of containing definite technical errors. But on closer scrutiny the
alleged errors turn out to be philosophical challenges to cherished assump-
231
 MATHEMATICS
tions about the nature of mathematics. On the other hand, while these chal­
lenges are ingenious and radical, they are controversial and often proble­
matic.
Mathematics provided Wittgenstein's gateway into philosophy, but he
quickly moved on to the nature of logic and representation. His treatment
of mathematics emerged relatively late (NB 17.8./21.11.16) and occupies
only two brief passages in the Tractatus's discussion of non-empirical would-
be propositions (TLP 6.02-6.031 & 6.2-6.241). Nevertheless, the Tractatus
contains profound objections to Frege's and Russell's logicism. Logicism is
the attempt to reduce mathematics to logic. It aims to provide mathematics
with a secure foundation, and to show against Kant that mathematical pro­
positions are not synthetic a priori but analytic, in that their proof relies
exclusively on logical axioms and definitions (Notation Pref., §13; Foundations
§§3-4). The concepts of mathematics can be defined in terms of logical con­
cepts; its propositions can be derived from logical principles through logical
deduction.
The Tractatus challenges logicism in several respects: (a) its SAYING/SHOWING
distinction rejects Russell's attempt to avoid the set-theoretic paradoxes
through the theory of types; (b) it challenges the axiomatic conception of
LOGIC according to which there are more- and less-fundamental necessary
truths (axioms and theorems respectively), and thereby the idea that deriving
mathematical propositions from logical axioms grounds them in something
more certain or evident; (c) it criticizes the logicist definition of NUMBERS, and
submits a constructivist alternative, according to which the natural numbers
represent stages in the execution of a logical operation.
Just as numerals do not stand for abstract objects, mathematical equations
do not say anything about a Platonist world. Rather, they equate signs
which are equivalent by virtue of rules governing reiterable operations (TLP
6.2ff.). Like the TAUTOLOGIES of logic, the equations of mathematics say noth­
ing about the world, but 'show' its 'logic', presumably because they display
the structure of truth-functional operations. However, while tautologies are
'senseless' propositions, equations are 'pseudo-propositions', on a par with
the nonsensical pronouncements of metaphysics (TLP 6.2-6.22). It may
seem that the reason for this mscrimination lies in the fact that the Tractatus
eliminates the iDENTiTY-sign from its ideal notation. But this is not the whole
story, since that ideal notation equally represents truth-functional relations
without LOGICAL CONSTANTS. The real difference is that tautologies are limit­
ing cases of meaningful empirical propositions. Equations are not, and
unlike vacuous tautologies, they seem to be saying something. However,
according to the saying/showing distinction, the 'identity of meaning' (here
taken in a non-Fregean sense) between, for example, the signs '2 x 2' and
'4', that is, the fact that they can be substituted for each other, cannot be
asserted by a meaningful proposition, it must be seen from the expressions
232
MATHEMATICS
themselves (TLP 6.23ff.), although only if they are properly analysed, which
in the case of complex equations would require substantial calculations.
'[I]n real life, a mathematical proposition is never what we want. Rather,
we make use of mathematical propositions only in inferences from proposi­
tions that do not belong to mathematics to others that likewise do not
belong to mathematics. (In philosophy the question, "What do we actually
use this word or this proposition for?" repeatedly leads to valuable insights.)'
(TLP 6.211). This passage contains the seeds of Wittgenstein's later account
of mathematical propositions. After 1929, Wittgenstein abandoned the
saying/showing distinction, and, as a result, no longer treated mathematical
equations as pseudo-propositions. Instead, he followed his own recommen­
dation and examined the role of mathematical propositions in empirical
reasoning. This sets his account apart from traditional discussions of neces­
sary truths, which are preoccupied with questions like 'What is the source of
necessary truths?' and 'How can we come to know them?' Wittgenstein, by
contrast, is concerned with the prior question of what it is for a proposition
to be necessarily true. And he answers that question by looking at how these
propositions are actually used, what their role is.
In doing so, he takes up a profound problem which had been ignored by
logicism, and which defies the Platonist picture according to which mathe­
matical propositions are truths about a separate ontological realm of abstract
entities, but which had been detected by Kant. Mathematical propositions
appear to be synthetic a priori; they do not rely on experience, but never­
theless seem to hold true of the objects of experience, that is, of the material
world rather than a Platonic kinterworld. Wittgenstein takes seriously the
empiricist position, explicit in Mill and implicit in Russell and Ramsey,
according to which mathematical propositions are well-confirmed truths
about the most pervasive aspects of material reality, because it is based on a
hard-nosed 'realism' which avoids both arcane abstract entities (Platonism)
and arcane mental structures (Kant's pure intuitions). But he rejects it,
because he recognizes that 'no experience will refute' mathematical proposi­
tions, if we put two apples into a bucket and add another two, but find only
three apples on emptying the bucket, we conclude not that, exceptionally,
2 + 2 = 3, but that one apple must have vanished (AWL 197; RFM 325).
2
We can use an equation like '25 = 625' descriptively, for the purpose of
predicting what result people will get when they square 25. But in fact we
use it normatively, to lay down what result people must get, if they have
squared 25: the result is a criterion for having performed the operation con­
cerned: if you do not get 625, you must have miscalculated, that is, you
have not squared 25. 'Calculation is not an experiment' (TLP 6.2331; see
AWL 185-91; RFM 221, 308-10, 318-19, 327-30, 359-63, 392-3).
This provides the key to Wittgenstein's own account. Mathematical pro­
positions describe neither abstract entities nor empirical reality, nor do they
233
 MATHEMATICS
reflect the transcendental workings of the mind. Their a priori status is due
to the fact that, in spite of their descriptive appearance, their role is a nor­
mative one: nothing which contravenes them counts as an intelligible
description of reality: 'There are 2 + 2, that is, 3 apples in the basket' is
nonsensical (RFM 363, 425, 431; LFM 55; see FORM OF REPRESENTATION).
Mathematical propositions are rules of GRAMMAR, 'paradigms' for the trans­
formation of empirical propositions. Arithmetic equations do not describe
relations between abstract entities, but are norms for describing the numbers
of objects in the empirical world, that is, substitution rules. '2 + 2 = 4' licen­
ses one to pass from 'There are two pairs of apples in the bucket' to 'There
are four apples in the bucket.' By the same token, an inequation like '4 > 3'
permits the characterizing of a quartet as larger in number than a trio, and
precludes 'This trio is larger in number than that quartet' (WVC 62, 153-7;
PR 143, 170, 249; PG 347; RFM 98-9, 163-4; MS123 98). Geometrical
propositions are rules for describing the shapes of and spatial relations
among objects, and for the use of words like 'length', 'equal length', etc.
They also set up ideals or norms for describing a measurement as accurate
(WVC 38, 61-3, 162-3; PR 216; LWL 8, 55; PG 319; RR 127; LFM 256;
PLP 44). 'The sum of the angles of a triangle is 180°' specifies that if figure
A is a triangle, its angles must add up to 180°.
The idea that mathematical propositions are norms of description cor­
recdy explains applied mathematics, by identifying the role of mathematical
propositions within empirical discourse. It should ensure Wittgenstein's place
in the philosophy of mathematics, even if his account of how we arrive at
such norms in pure mathematics is inadequate (see MATHEMATICAL PROOF). It
separates his position from the established schools of twentieth-century
philosophy of mathematics, which are united by the idea that mathematical
propositions refer to some kind of reality, whether physical signs (formalism),
mental processes (intuitionism) or abstract entities (logicism).
Logicism Like logic, mathematics moves within the rules of our language,
and is hence unassailable by experience. Nevertheless, Wittgenstein retained
the idea that there is a difference between the tautologies of logic, which say
nothing, and hence cannot express a rule, and mathematical propositions,
which themselves express rules (RFM 98-9; WVC 35, 106-7, 218-19; PR
126; AWL 146-8; LFM 272-85). The logical positivists ignored this distinc­
tion, and hence believed that Wittgenstein's account merely added to logi­
cism the idea that mathematical propositions are tautologies.
Intuitionism Wittgenstein, influenced by Schopenhauer and Spengler, shared
the anti-intellectualist oudook of Brouwer's intuitionism, and the idea that
mathematics rests on human activity. But he rejected the idea that this
activity is non-linguistic and mental, and rests on a 'basic intuition'. He
234
MATHEMATICS
agreed with Brouwer that the law of excluded middle does not apply to
'Four consecutive 7's occur in the expansion of But his point was that
there is no such thing as the expansion of it - an actual infinity - only an
unlimited technique for expanding 7t, and hence expansions of n up to n
places (WVC 71-3; PR 146-9; AWL 140, 189-201; PG 451-80; RFM
266-79; PI §§352, 516; PLP 391-6). Moreover, he rejected Brouwer's and
Weyl's idea that such sentences are meaningful, yet undecidable because
they transcend our recognitional powers. Instead, he argued, in a VERIFICA-
TIONIST vein, that a mathematical proposition which is undecidable in princi­
ple does not have a third truth-value (undecidable) but is senseless, and that
the law of excluded middle partly defines what we mean by a PROPOSITION
(PR 176, 210; AWL 139-40; PG 458; LFM 237; PI §136). However, if
Wittgenstein is right that we are dealing with a FAMILY-RESEMBLANCE concept
here, this claim might have to be restricted, for example to propositions of
mathematics or of the predicate calculus (see BiPOLARrrY).
Formalism Unlike some nominalists or formalists, Wittgenstein is not com­
mitted to the claim that mathematical propositions are really about signs:
'2 + 2 = 4 ' is neither about signs (inscriptions or sounds), nor about how
people use signs. Yet, although it is not a meta-linguistic statement, it is used
as a rule for the use of signs. Wittgenstein tries to eschew both formalism
and Platonism by insisting that what distinguishes a mathematical symbol
from an empty sign, just as what distinguishes a chess-piece from a piece of
wood, is not that it describes abstract entities and relations, but that it has a
rule-guided use within our linguistic practices (WVC 103-5; LFM 112;
RFM 243; RR 128; see LANGUAGE-GAME). This by itself does not set his
position apart from those formalists who claim that mathematics is a rule-
guided game with signs. However, for Wittgenstein 'it is essential to mathe­
matics that its signs are also employed in mufti. It is the use qutside mathe­
matics, and so the meaning of the signs, that makes the sign-game into
mathematics' (RFM 257, see 232, 258-60, 295, 376). This does not mean
that all parts of mathematics must have direct empirical application, but
only that those which do not must be connected with parts that do. There
is no pure mathematics without some applied mathematics. Mathematics
would be just a game if it did not play a role widuh our empirical reasoning.
Wittgenstein not only disagrees with these different schools, but also ques­
tions the whole enterprise to which they are alternative contributions,
namely of providing mathematics with secure foundations. He makes two
basic points. One is that the attempts to ground mathematics, and in parti­
cular Hilbert's META-MATHEMATICS, fail, since they simply produce further
mathematical calculi. The other is that the fear of the sceptical threat posed
by CONTRADICTIONS and antinomies of the kind Russell detected in Frege's
235
r
MEANING
system is a 'superstition' (WVC 196; RFM 120-2). It cannot be overcome,
by constructing logical symbolisms - this is the 'disastrous intrusion' of logic
into mathematics - but only dissolved, by philosophical clarification (RFM
281, 300).
Wittgenstein distinguishes sharply between mathematics, which changes
our conceptual scheme by deriving new norms of representation (e.g., equa­
tions), and philosophy, which simply describes the evolving conceptual
scheme. According to Philosophical Investigations §§124-5, philosophy 'leaves
mathematics as it is'. It is concerned not with the technical cogency of the
calculations and proofs, but only with the 'prose' with which mathematicians
surround them, the philosophical descriptions they give of their significance
(WVC 149; PG 369, 396; RFM 142; LFM 13-14). However, in other pas­
sages Wittgenstein acknowledges that this distinction between mathematical
equations and philosophical prose is artificial. Without their prose context,
many proofs in mathematical logic and set theory would be mere games
with symbols. If Wittgenstein is right that this context is bedevilled by meta­
physical confusions, this may not change the proofs, but it should change
our 'attitude to contradiction and to consistency proofs'. It should make
mathematicians 'abandon' as uninteresting, for example, transfinite set
theory, and should slow down the growth of new formal systems (RFM 213;
CV 1-2; LFM 103; PG 381-2). Wittgenstein's much-discussed 'non-revision­
ism' boils down to the idea that while technical advances in mathematical
logic may create new philosophical problems they cannot solve them, since
these problems require conceptual clarification (PI §125; RFM 388).
Another contribution to the philosophy of mathematics is Wittgenstein's
anthropological perspective on mathematics as part of the natural history of
mankind, and the idea that mathematics is a family of activities for a family
of purposes (RFM 92-3, 176, 182, 399).
meaning (Bedeutung) This concept occupies a central role in Wittgenstein's
work, because of his abiding conviction that philosophical problems are
rooted in language. His later work invokes and elucidates the everyday
notion of linguistic meaning (see USE). His early discussion is a metaphysical
reflection on the nature of symbolic representation, and evolves from a
technical dichotomy between 'sense' and 'meaning' adapted from Frege.
Frege was not concerned with all aspects of the meaning of expressions, for
example not with their 'colouring', the mental associations they evoke, but
only with those which bear on the validity of arguments in which they
occur, their logical 'content'. In his mature system he distinguished two
aspects of content: sense (Sinn) and meaning (Bedeutung). In an ideal lan­
guage every sentence expresses a sense, the thought (what is judged), and
refers to a meaning or referent, a truth-value, the True or the False. It
expresses a thought by presenting a truth-value as the value of a function
236
MEANING
for an argument. Each significant constituent of a sentence also expresses a
sense and has a referent. Proper names express a sense and refer to an
object, concept-words express a sense and refer to a concept. This distinc­
tion explains both how an expression may fail to refer without being sense­
less, and the non-trivial nature of IDENTITY statements like 'The morning
star is the evening star': although the two expressions mean the same
object, their sense or 'mode of presentation' differs ('Sense' 25-36; Laws I
§2; Correspondence 63; Notation §§2-8).
Initially, Wittgenstein accepted the idea that propositions have a meaning
(Bedeutung), while rejecting other aspects of Frege's position (NL 94-104; NM
112-13). Neither the sense nor the meaning of a proposition is an object.
The meaning of 'p' is not its truth-value, but the fact that corresponds to it
in reality, that 'p' if it is true, that ~p if it is false. PROPOSITIONS differ from
names. They are BIPOLAR - capable of being true and capable of being false
- which is precisely to say that they have a sense. To understand a name is
to know what it refers to, but to understand a proposition one need not
know whether it is true or false, but only what would be the case in either
event.
The Notebooks sharpen this contrast by gradually abandoning the idea that
propositions have a meaning (NB 20.9./2.10./26.10./2.11.14). As a result,
the Tractatus maintains that names have a meaning but no sense, while pro­
positions have a sense but no meaning (TLP 3.142, 3.203, 3.3). (The Tracta­
tus also employs the terms 'meaning' and 'sense' non-technically (e.g. TLP
5.02, 5.451, 6.521), a fact which has misled some commentators.) The sense
of a proposition is 'what it represents', namely a possible 'state of affairs' or
'situation', an arrangement of objects which may or may not obtain,
depending on whether the proposition is true or false. The proposition shows
its sense, 'how things stand if it is true. And it says that they do so stand'
(TLP 4.022, see 2.20Iff.; see also SAYING/SHOWING). The sense of a proposition
is neither an object that corresponds to it, a Fregean thought, nor the mode
of presentation of a truth-value, but a possibility, a potential combination of
objects which need not be realized.
Sense antecedes the facts: in order to decide whether a proposition is true,
its sense must be determined; to understand its sense we need not know its
truth-value, but only 'what is the case if it is true' (TLP 4.024, 4.061-4.063;
NB 24.10.14; Laws I §32). This idea goes back to Frege, and lies at the heart
of modern truth-conditional semantics. The sense of a truth-function of 'p' is
a function of the sense of '/>'. Negation, for example, reverses the sense of the
proposition. '/>' and '~p' have 'opposite sense', even though one and the
same reality corresponds to them: a single fact verifies one of them and falsi­
fies the other (TLP 4.0621, 5.122, 5.2341; NL 95, 105; NB 6.5.15). The
sense of a 'molecular' proposition is given by its 'truth-conditions', that is, by
determining for each of the possible combinations of truth-values among its
237
 MEANING
constituents (elementary propositions) whether it comes out as true or as false
c
in a TRUTH-TABLE: '/>. q' is true if both '/>' and q' are true, false if either or
both are false (TLP 4.431). An ELEMENTARY PROPOSITION cannot have truth-
conditions in this sense. But it can have 'truth-grounds', and to understand
them is to know what is the case if it is true (TLP 5.101-5.121). To know
what is the case if a molecular proposition is true is to know what elemen­
tary propositions make it true, to know what is the case if an elementary
proposition is true is to know what possible combination of objects corre­
sponds to the way its elements are combined. There is a condition which an
elementary proposition must satisfy to be true, namely that it depicts objects
as being combined in the way they actually are.
The sense of an elementary proposition is determined by the meanings of
its simple 'constituents', that is, NAMES (TLP 3.318, 4.026f). In order to
understand an elementary proposition, we need to know what OBJECTS its
constituent names stand for. 'A name means an object. The object is its
meaning' (TLP 3.203, 3.22). This is a straightforward version of the AUGUS­
TINIAN PICTURE OF LANGUAGE which Wittgenstein later rejected: the meaning of
a name is the object which it 'represents' (vertreten); meaning (bedeuten) is a
one-to-one relationship between names and objects. The Tractatus's position
here has been defended on the grounds that it employs a technical notion,
and that the Bedeutung of a name is its semantical role, the contribution it
makes to the sense of a proposition. However, that technical use precisely
identifies the meaning of a word with what it stands for. And a name con­
tributes to the sense of a proposition by 'going proxy' for an object. The
very 'possibility of propositions' is based on this relation: unless a name has
been associated with an object, propositions in which it occurs will lack a
sense (TLP 4.031 If., 5.473, 6.53). This is no aberration, but essential to the
PICTURE THEORY: a proposition can be false yet have a sense only because,
although no fact corresponds to it as a whole, it consists of elements which
are correlated with elements of reality.
It is more plausible to defend the Tractatus by claiming that, as in Philoso­
phical Investigations, the meaning of names is determined by their use: what
objects they stand for depends on their logical syntax, the way they behave
in propositions. The Tractatus indeed condones a version of 'Occam's
Razor': signs which have no 'logico-syntactical employment', no role in
representing reality, are meaningless; two signs with the same employment
mean the same (TLP 3.326ff.; NB 23.4.15). Moreover, it is correct that one
can learn the meaning of a name from its use in propositions. But for the
propositional sign to have a sense in the first place it must be projected onto
reality by the mind. And although Wittgenstein speaks of the METHOD OF
PROJECTION as the 'application of the propositional sign', this in turn is identi­
fied with a mental activity, 'the dunking of its sense' (PT 3.13; TLP 3.11).
The speaker thinks a mere propositional SIGN onto a possible state of affairs.
238
MEANING-BODY
And to do that, he must correlate its elements with the elements of the state
of affairs depicted.
The Tractatus's discussion of meaning and sense was an important step in
the development of semantics. It features important insights: the denial that
LOGICAL CONSTANTS stand for something, and the contrast between proposi­
tions and names; and prefigures others: the importance of linguistic use. But
it remains wedded to mistakes: a referential conception of meaning, and the
idea that a proposition's sense must be DETERMINATE.
meaning-body (Bedeutungskdrper) Wittgenstein uses this term to character­
ize the idea that behind each sign there is a non-linguistic entity, its mean­
ing, which determines how it is to be used correctly. According to this view,
a word is analogous to a single painted surface of an otherwise invisible
glass-body with a certain geometrical shape (e.g., a cube or pyramid). The
combinatorial possibilities of the visible surface depend on the shape of the
body behind it. Similarly, grammatical rules are seen as the geometry of
meaning-bodies. We can derive the rules for the use of a word from its
meaning, since the latter is a (concrete, abstract or mental) entity which
determines the combinatorial possibilities of the word (PG 54; AWL 50-1;
PLP 234-7). Grammatical rules are not AUTONOMOUS, but responsible to the
'true' or 'real' meaning of the sign involved, something outside language
which can be discovered through LOGICAL ANALYSIS.
Such a view is prominent in Frege, who thought that he had for the first
time revealed the true meaning of number-words, and insisted, against the
formalists, that the rules for the use of mathematical symbols must 'follow
from what they stand for', their meanings (Foundations Introd.; Laws II §§91,
136). It can also be detected in the early Wittgenstein, who thought that the
identity '—/>=/>' mirrors 'the fact that double negation is an affirmation' (NB
4.12.14), which in turn is an aspect of the essential BIPOLARITY of the propo­
sition. On the other hand, one of the ideas behind the Tractatus's SAYING/
SHOWING distinction was that we cannot derive th& rules governing the use of
a sign from its meaning, since the sign does not have a meaning in advance
of these rules.
Wittgenstein later directed this idea against Frege's Platonism, the Tracta­
tus's metaphysics of symbolism and the mentalism ofJames (Psychology I 245-
6; see also Analysis 252), for whom the meaning of a logical term like 'not' is
a feeling (e.g., of rejection) which we associate with it (PG 58; BT 42). His
arguments also threaten the seemingly innocuous claim of logical positivism
that the truth of tautologies follows from the TRUTH-TABULAR definitions of
the logical connectives, and the attempt of model theory to show that our
rules of inference follow from the semantical definitions of the logical con­
stants. All these positions derive what for Wittgenstein are GRAMMATICAL
propositions or rules from meanings.
239
 MEANING-BODY
Against this idea, Wittgenstein adduces two interrelated arguments. Firsdy,
while a rule can logically follow from another rule (that 'Betty' is written
with a capital 'B' follows from the rule that all proper names are written
with a capital), it is unclear how it could follow from a meaning (PLP 236).
Secondly, necessary propositions do not follow from the meanings of signs or
from linguistic conventions, they partly constitute them. For to abandon a
necessary proposition is to change the meanings of at least some of its con­
stituent signs.
Grammar is not accountable to any reality. It is grammatical rules that
determine meaning (constitute it) and so they themselves are not answer­
able to any meaning and to that extent are arbitrary. (PG 184, see 52-3,
243-6; AWL 4; RFM 42; LSD 20).
Rules of inference, for example, determine the meaning of the logical con­
stants, rather than proceed from them. Whether a specific transformation of
symbols is licensed or not is one aspect of the correct use and hence of the
=
meaning of the terms involved. That we use ' — p p ' as a rule of inference
contributes to the meaning of '"'. Without that rule, the sign would not
have the meaning it has. And if the rule were changed, if we accepted
=
instead '—p ~p\ the meaning of '"' would change correspondingly.
Accordingly, the rules of inference cannot correspond or fail to correspond
to the meaning of, for example, negation. Someone who passes let us say
from '—p' to '~p' does not follow a false rule of negation, but has given a
different meaning to ' - ' (PI 147n; RFM 398).
There are three problems with these arguments. One is mentioned by
Wittgenstein himself, namely that questions of identity and difference of
meaning are more complex than they allow (PI §§547-59). If two people use
'not' in the same way except that one of them uses double negation empha­
tically and the other as equivalent to assertion, we would not say that they
employ 'two species of negation'. For we would not say that 'not' means
something different for each of them in 'Do not enter the room.' On the
other hand, we would say that it does mean something different in T ain't
done nothing' (RFM 104). Questions of synonymy are context-dependent.
Secondly, to say that '—/>=/>' follows from the truth-tabular definition of'"'
can be understood innocuously as the contrapositive of Wittgenstein's own
claim. From Wittgenstein's claim that if we alter the rule we alter the mean­
ing, it follows that if we do not alter the meaning we get the rule. Thirdly,
=
although we could use according to either ' — p p ' or ' — p = ~ p \ it
would be inconsistent to combine our truth-tabular explanation of it with the
second rule. For in that case we would say that the truth-table has been
misunderstood. By Wittgenstein's own lights, the truth-tabular explanation is
1
a rule, and to accept '~~p= -p is a criterion for having misunderstood that
240
MEMORY
rule, because one is not applying to '~p' the same operation (that of rever­
sing the truth-value) that has been applied to '/>'.
To the last point, Wittgenstein replies 'Who says what "the same thing"
is' (LFM 180, see 81-2; RFM 102-6; FW 57-8). What he has in mind is
that the rule follows from the explanation only if it is understood that in the
truth-table the place of 'p' can be taken by '~/S' (that we apply negation as
in '"("/>)' and not as in '(—pf). Accordingly, ' — p = p' is not determined by
the truth-table definition alone, but only in conjunction with this second
rule. Since there is no comparable rule in natural languages, nothing deter­
mines how to understand T ain't done nothing' (LFM 184). But this leaves
the first two objections. It seems that 'The rules determine the meaning' is
as wrong as 'The meaning determines the rules.' Understanding the truth-
tabular explanation and acknowledging ' — p p ' are simply INTERNALLY RELA­
=
TED aspects of one and the same practice of using '-'. The truth-table would
mean something different in a practice in which ~~p=p'i
were rejected.
They are simply two different rules of our practice and both of them are
constitutive of that practice.
This leaves intact the original case against meaning-bodies. Signs as such
don't have meanings. There are no entities from which the use of a sign
'flows', or which force us to use, for example, the truth-table one way or
another. We give meaning to signs by explaining and using them in a parti­
cular way; and by employing them differendy we can change their meanings
(BB 27-8; AWL 50-1, 131-2; see RULE-FOLLOWING). The rules we adopt are
neither correct nor incorrect. This chimes in with Quine's claim that the
logical positivists' idea that necessary truths are true by virtue of meaning is
based on the 'myth of the museum', the idea that there are abstract or
mental entities - meanings or logical forms - which force us to hold on to a
certain form of words come what may. But unlike Frege and the Tractatus,
the logical positivists and model theory could replace talk of meanings by
talk of explanations. Yet, if rules and explanations are clifferent aspects of
one and the same practice, one cannot understand the explanation and then
see what rules follow from it. Rather, to understand the explanation is to
acknowledge the rules.
m e m o r y The traditional view is that memory is a storage system, a piece
of wax (Plato) or a store-house of ideas (Locke) which contains previous
impressions or experiences, or at least their traces (Aristode). According to
this picture, when I remember a thing or an event X, I retrieve a mental
image of X and parade it before my mind's eye; when I recognize X, I
notice that my current impression of X fits with a mental image derived
from a previous experience of X. The difference between a current and a
remembered experience is then held to lie either in the greater vivacity of
241
 MEMORY
the former (Hume, A Treatise of Human Nature 1.1 .v), or in a feeling of famil­
iarity accompanying the latter (James, Psychology I ch. XVI; Russell, Analysis
ch. LX).
The later Wittgenstein condemned this conception of memory and recog­
nition as 'primitive' (BB 165). His reflections were inspired by James and
Russell, although Augustine's Confessions (ch. X) may also have played a role.
For one thing, Wittgenstein rejected the idea that memory essentially
involves mental images. Although mental images may accompany my
remembering X, they are neither necessary nor sufficient. Moreover, even in
cases in which mental images cross my mind, I do not read off what hap­
pened from the images. When I remember, say, having wished to O or
having meant so-and-so, or what a perfect number is, I do not and cannot
read off what I remember from any mental image (PI §§645—51; RPP I
§468).
Even if X is something that can be pictured, having a mental image of X
does not guarantee remembering. As imagists like James realized, the image
would still have to be certified as a representation of something past. Pace
James, however, this cannot be explained by a special feeling of familiarity
or 'pastness'. Firsdy, I would in turn have to recognize, that is, remember
this feeling. Secondly, I can connect a feeling with the past only if I discover
that it is regularly associated with memories as opposed to other kinds of
experience; but I would need to rely on my memory to make this correla­
tion. It takes memory to tell me whether what I experience is the past.
Finally, connecting such a feeling with the past presupposes having a con­
cept of the past, but that concept is in turn learnt by remembering. More
generally, remembering X cannot be explained as the occurrence of a
'memory-experience', since memory is presupposed in linking experiences to
the past. Remembering has 'no experiential content', that is, nothing that
happens while I remember is the remembering (PI §§595-6, II 231; LW I
§837; Z §662). Although characteristic mental experiences or processes may
accompany remembering, they do not constitute^ it. This reasoning underlies
Wittgenstein's denial that remembering is a mental process or experience,
and that there is a uniform connection between remembering and what is
remembered (PI §§305-8; PG 79-80).
Even if one abandons the imagist conception, one may agree with Aris­
totle (On Memory 450a—b) that I can remember X only if the original experi­
ence of X has left some physiological trace in me. This idea of memory
traces was accepted, for example, by James, and developed by Kohler, who
held that the brain must contain a physiological record which is isomorphic
with the recorded experience (Gestalt 210-11). Wittgenstein attacked Kohler's
reasoning (RPP I §§220, 903-9; Z §§608-13). For one thing, he pointed out
that when we remember we do not read the past event off a neurophysiol­
ogical trace: unlike a written record, such a trace has no symbolic content.
242
METALOGIC/-MATHEMATICS/-PHILOSOPHY
For another, he questioned Kohler's plausible assumption that the remem­
bered events cannot have a present effect - the remembering - unless they
continued to exist in some way. According to Wittgenstein, there might be a
psychological regularity, a causal relation between the experience and the
remembering, to 'which no physiological regularity' corresponds. This is to
deny that there must be a psychophysical parallelism and thereby, by Witt­
genstein's own admission, to challenge our conceptions of CAUSALITY.
On the other hand, Wittgenstein tacidy accepts the idea that the connec­
tion between the remembered event and the remembering must be causal to
begin with. Although this assumption is shared by the currendy dominant
causal theory of memory, it can be questioned on Wittgensteinian grounds.
I remember X now because I experienced X earlier, but the 'because' here
seems a GRAMMATICAL one: it is part of our concept of memory that unless I
had experienced X, I could not possibly remember X. By contrast, that
there is a causal connection between experience and remembering seems to
be a scientific discovery.
Wittgenstein disputes the view of recognition as the matching of an object
or a current sense-impression with a stored mental image (PI §§596-610;
PG 179-82; BB 84-8, 165; RPP I §1041). Firsdy, recognition need not
involve a mental image of what is recognized. Secondly, even if a mental
image of X does accompany recognizing X, it cannot explain it, for one
would in turn have to recognize that the image is an image of X. Thirdly, it
is wrong to hold that a process of recognition takes place whenever we per­
ceive familiar things: when I enter my study, I neither recognize my desk,
nor fail to recognize it.
This last claim has been attacked along Gricean lines: the fact that we do
not say that I recognized the desk does not imply that I did not. But it is
incumbent upon Griceans to show that the fact that we would not speak of
recognition in such cases is due to general pragmatic maxims rather than to
semantic features pertaining specifically to the term 'recognize' (see WILL).
Wittgenstein's attack on the idea that stored representations are essential
to memory and recognition pertains not just to the imagist tradition, but
also to the idea of representations in the brain which ranges from Kohler to
Marr. His claim that nothing need have occurred when I remember X was
elaborated by Malcolm, who holds that to remember X is simply to have
experienced or learnt X, and not to have forgotten it, and that the causal
connection between experience and remembering is a contingent fact rather
than part of the concept of remembering.
metalogic/-mathematics/-philosophy Wittgenstein declares the rejec­
tion of such 'meta-disciplines' to be a 'leading principle' of his philosophy
(PG 116). It is part of his anti-foundationalist conception of philosophy and
is directed against the idea that philosophy is needed to either justify or
243
 METALOGIC/-MATHEMATICS/-PHILOSOPHY
explain our ordinary (i.e., non-philosophical) uses of language. The idea that
we cannot know anything unless we have answered the question 'What is
knowledge?' is as absurd as thinking that we cannot spell at all unless we
have completed a meta-investigation into the spelling of 'spelling' (PI §121;
TS219 10). By the same token, there are no 'essential problems in philosophy'
which must be solved before anything else can be done. While different pro­
blems have a special importance at particular stages in the history of philo­
sophy, none of them are intrinsically fundamental (BT 407; CV 10; RPP I
§1000). This is the 'real discovery' in philosophy, because it 'makes me cap­
able of stopping doing philosophy when I want to' without leaving every­
thing 'hanging in the air' (PI §133; BT 431-2). There are several aspects to
this idea.
Meta-mathematics Hilbert used this term to refer to his programme of estab­
lishing the consistency of mathematics by making mathematical proofs the
topic of another mathematical calculus. Wittgenstein's rejection of meta-
mathematics is directed not just at Hilbert's programme (WVC 120-1, 133-
6; PR 180), but more generally at any attempt at providing foundations for
mathematics, including the original logicist programme of reducing it to
logic (AWL 12-13, 68; PG 296-8; LFM 260-2, 271-2). Wittgenstein rejects
this idea through a simple regress argument. Both meta-mathematics arid
the logical systems of Frege and Russell are themselves nothing but further
calculi,- more MATHEMATICS 'in disguise'. Indeed, they are less basic than stan­
dard arithmetic, since they are remote from our mathematical practices and
presuppose a grasp of standard arithmetic. 'They are no more the founda­
tion of mathematics for us than the painted rock is the support of a painted
tower' (RFM 378).
Metaphilosophy Nowadays this simply refers to philosophical methodology.
The term was introduced by Lazerowitz, to indicate a non-philosophical dis­
cipline which explains the nature of philosophy by combining Wittgen-
steinian and Freudian ideas. Ironically, Wittgenstein himself took the
traditional line that the nature of philosophy is itself a philosophical pro­
blem, and explicitly rejected the idea of metaphilosophy: 'One might think:
if philosophy speaks of the use of the word "philosophy" there must be a
second-order philosophy. But it is not so: it is, rather, like the case of ortho­
graphy, which deals with the word "orthography" among others without
then being second-order' (PI §121; see LSP 25). He relates this to the idea
that ordinary language, including ordinary, non-philosophical employments
of specialized languages, is fundamental to philosophy. Philosophical pro­
blems concern expressions which already have a non-philosophical use (RPP
I §550). This would be granted by ideal language philosophers like Carnap.
However, they blame philosophical problems on the ambiguity and vague-
244
METALOG1C/-MATHEMATICS/-PHILOSOPHY
ness of ordinary language, and try to resolve them through the introduction
of artificial calculi in which these problems cannot be formulated. But if the
problems arise from ordinary concepts, their resolution must clarify those
concepts. As Strawson put it: artificial concepts can cast light on these diffi­
culties only if their relation to our ordinary concepts is understood, which
presupposes an accurate understanding of the latter. This by itself will
achieve the desired resolution, if Wittgenstein is right in claiming that philo­
sophical problems arise not out of deficiencies of ordinary language, but out
of its misuse or misconstrual in philosophical reflection - when 'language
idles' (PI §§38, 89).
Introducing a new notation may remove possible sources of philosophical
error: we may curb the temptation for Hegelian confusions about 'identity
in difference' by adopting a notation which replaces 'is' by either '=' or 'e'
(PI §90; TS220 §99; see IDENTITY). But this presupposes that our 'is' in fact
expresses both identity and predication. Introducing new grammatical rules
plays a (limited) role in clarifying the old ones. But unless we have achieved
the latter, we will not b e able to cope with the new problems which any
novel notation will create. New notations, whether formal languages or fic­
tional LANGUAGE-GAMES, are useful mainly as 'objects of comparison . . . to throw
light on the facts of our language by way not only of similarities, but also of
Dissimilarities' (PI §130, see also §§2-64).
Metalogic Carnap's Logical Syntax of Language ascribes the origin of this term
to the Warsaw logicians. At present, it is used to refer to second-order
reflections about logic (e.g., proofs of soundness and completeness). Wittgen­
stein himself uses this term - mainly in the 'Big Typescript' (BT 3, 16, 205,
282, 285-6). Logic determines what is necessary, but there is no metalogic
which makes logic necessary. We cannot step behind the distinction between
sense and nonsense drawn by logic (PG 126—7). Wittgenstein also denies
that there are metalogical concepts. This has been presented as directed
against the view that psychological concepts like understanding or meaning
denote mental phenomena which give language its meaning. While this
interpretation fits some passages (Z §284; BT 1; MSI 10 189-91; MSI 16
16), it is too narrow. Wittgenstein uses the term 'metalogical' for non-
psychological concepts (BT 412; PG 101) and maintains that all concepts
which philosophy uses in describing ordinary language are themselves ordin­
ary.
When I talk about language . . . I must speak the language of every day.
Is this language somehow too coarse and material for what we want to
say? Then how is another one to be constructed? - And how strange that we
should be able to do anything at all with the one we have! In giving ex­
planations I already have to use language full-blown (not some sort of
preparatory, provisional one) . . . (PI §120)
245
 METALOGIC/-MATHEMATICS/-PHILOSOPHY
This is directed firsdy against the Tractatus, which had already insisted that
philosophy elucidates ordinary language, but had assigned an extra-ordinary
status to the concepts used to do so. According to the SAVING/SHOWING dis­
tinction, 'proposition', 'name', 'function', etc., are 'formal concepts' which
cannot even be meaningfully employed. The price formal concepts pay for
being taken off the index in the later philosophy, is that their legitimate use
is as 'low' and 'homely' as that of ordinary 'material' concepts (PI §§97,
108-9; PG 121). It is also directed against James's idea that ordinary con­
cepts are too coarse to describe mental phenomena, partly because the latter
slip by too quickly (PI §§436, 610; PG 169; Psychokgy I 195, 251), and
against the idea that a phenomenologicaT language referring to sense-data
is semanticalry primary (see VERIFICATIONISM).
Wittgenstein supports his view that there are no more-fundamental or more-
refined artificial concepts for philosophy to rely on with a regress argument.
In elucidating ordinary concepts (e.g., 'red', T), philosophy may use techni­
cal terms like 'colour-predicate', 'indexical' or 'language-game', as well as
terms like 'foundations' or 'philosophy'. But if the terms employed in philo­
sophical clarification were part of a meta-symbolism, there would be a need
for a further clarification in yet another language, and so on. We would end
up with an 'irifinite hierarchy' (LFM 14) of meta-languages, the equivalent
of the regress of justification we encountered with respect to meta-mathema­
tical calculi. Artificial languages cannot be constructed in a vacuum. At least
some of their expressions have to be explained in terms which are already
familiar, ultimately those of ordinary language, 'which must speak for itselF
(BT 1; PG 40; PI §§5-6; Z §419). With respect to many purposes, ordinary
language is inferior to technical idioms. But it is the semantic bedrock:
through acquiring ordinary language we acquire the ability to learn and
explain new and technical terms. There is no semantic exit from this lan­
guage, either upwards into a hierarchy of meta-languages, or downwards to
reality (see OSTENSIVE DEFINITION). We come to it not through another lan­
guage, but through training in basic linguistic skills (see EXPLANATION).
It has been claimed that for Wittgenstein grammar is flat: there are no
rules or concepts which are more fundamental than others. Wittgenstein's
rejection of metalogic actually suggests that the concepts of ordinary lan­
guage are fundamental in that we cannot 'get behind' them (PG 244). We
cannot describe our practice of following rules in more basic terms than the
rule-formulations of the participants. Those who do not understand those
formulations cannot be enlightened through a 'preparatory' language, but
can only be taught to participate (RFM 330, 392-3; Z §§310-19).
At the same time, Wittgenstein did cast doubt on the idea that there are
'categories', general concepts which signify basic structures of language and
provide the sole topic of philosophy. His reflections on coLOUR-terms show
246
METHOD OF PROJECTION
that words belonging to the same category do not share all their combina­
torial possibilities; thereby prefiguring later objections to Ryle's definition of
categories as classes of expressions which can be substituted for each other
sahia signifkatione. Nevertheless, Wittgenstein shared Ryle's aspirations: the
grammatical differences he sought to teach us are category differences in a
loose sense (RPP I §793; RPP II §§7, 690; Z §86). Moreover, his idea that
concepts like 'thinking', 'inferring', etc., impose conceptual limits on alter­
native grammars parallels the Kantian idea that categorial concepts are
constitutive of the concept of experience or of a conceptual scheme (see
AUTONOMY OF LANGUAGE).
However, Wittgenstein's attack on metalogical concepts rightly points out
that categorial concepts like 'experience', 'act', 'event', 'state' or 'process' are
not semantically prior: understanding them is not a precondition of under­
standing other concepts. Rather, they are devised in philosophy, in order to
characterize the logical role of classes of non-categorial terms (see PLP 103-
6). Moreover, categorial terms do not provide a sharply defined basis for
philosophy. 'These extremely general terms have an extremely blurred
meaning. They relate in practice to innumerable special cases, but that does
not make them any solider; no, it rather makes them more fluid' (RPP I
§648). It is precisely for this reason that they are so prone to cause philo­
sophical confusion.
Finally, Wittgenstein explicitiy renounced the idea, subsequently cham­
pioned by Dummett, that the philosophy of language is the foundation of
philosophy. We do not have to clarify concepts like 'language', 'meaning' or
'grammar' before we can clarify, for example, ethical concepts. For we can
describe the grammar of 'virtuous' or 'duty' without relying on a description
of the grammar of 'meaning'. But Wittgenstein is committed to regarding
some concepts as fundamental in a methodological sense, since he main­
tains, for example, that PHILOSOPHICAL problems are based on conceptual con­
fusions arising from misconstruing the meaning of words (LWL 61; M 51,
114; AWL 31).
m e t h o d of projection According to the PICTURE THEORY, a proposition
can depict a state of affairs only if its elements, NAMES, 'correspond' to, that is,
'stand' or 'go proxy for' (vertreten), the elements of the latter, OBJECTS (TLP
2.13f., 3.22, 4.031 If.). The 'correlations' between the elements of the picture
(thought, proposition) and the elements of the situation it represents are the
'pictorial relation' (abbildende Beziehung). These correlations are like 'feelers'
extending from the picture's elements, through which the picture itself
reaches right out to reality, that is, depicts a particular combination of objects
(TLP 2.1513ff). Wittgenstein also uses the term 'pictorial relation' for the
relation which obtains between picture and situation as a whole rather than
between their elements (TLP 4.014). In this use it seems equivalent to a
247
 METHOD OF PROJECTION
'method of projection', 'comparison', or 'depiction' (Prqjektions-/Vergleichs-/
Abbildungsmethode), or a 'manner of representation' (Barstellungsweise). Earlier, he
had contrasted method of projection and pictorial relation (NM 112). Even if
the elements of picture and situation have been correlated, it remains to be
determined which relations between the names are part of the picture's
'structure', that is, have symbolic significance in that they determine what the
proposition depicts. Equally, the fact that its elements are related in specified
ways (that it has a certain 'structure') only depicts a specific state of affairs if
these elements stand for specific 'things'. Accordingly, a picture consists of
structure plus pictorial relation; that is, of two relations, one between its ele­
ments, and one between the latter and reality. We can represent a specific
accident (which may or may not have occurred) with the aid of toy cars and
dolls only if we lay down both what toy corresponds to what actual thing,
and which relations between the toys represent actual relations between
objects (e.g., their spatial relations, but not those between their weights). In
subsequent writings, 'method of projection' includes both structure and pic­
torial relation, that is, eveiything required to compare a propositional sign
with a specific situation (NB 30.10.-1.11.14; TLP 3.11-3.13). The idea is
inspired by geometrical projection, which includes eveiything needed to
transform one figure (the proposition) into another (the depicted situation).
'A proposition includes all that the projection includes, but not what is
projected. Therefore, though what is projected is not itself included, its pos­
sibility is. A proposition, therefore, does not actually contain its sense, but
does contain the possibility of expressing it . . . A proposition contains the
form, but not the content, of its sense' (TLP 3.13, see 3.34). The proposition
does not 'contain its sense', the possible state of affairs, firstly because a con­
figuration of signs cannot contain the configuration of things which it repre­
sents, and secondly because, if the proposition is false, there will be no such
configuration to contain. What strictly speaking contains the 'possibility of
expressing' the sense is not the proposition, which does express it, but the
'propositional sign'. It does so because it shares a LOGICAL FORM with the
situation it depicts, it has the same logico-mathematical multiplicity (TLP
4.04) according to the conventions of LOGICAL SYNTAX.
These conventions determine only the combinatorial possibilities of
names, and thereby the logical form of the propositional sign. But the SIGN
as such does not depict; to become a symbol it must be given a content
through a method of projection. The method of projection is the 'applica­
tion of the propositional sign'. Correlating signs and reality is something we
do. This prefigures the later view that what endows signs with meaning is
not a correlated entity but their USE. Alas, the early Wittgenstein gives this
idea a mentalist gloss: the application of the propositional sign, and hence
'the method of projection', is 'to think the sense of the proposition' (das
Denken des SatzSinnes) (TIP 3.11; PT 3.12f). When we use a propositional
248
METHOD OF PROJECTION
sign with understanding, as a model of reality, we have to think its sense
into it, that is, we have to think of the situation depicted. Consequendy, a
continuous process of diinking and meaning accompanies and underlies
every meaningful use of signs. While diinking is a process, a thought is not.
Nor is it an abstract entity, as in Frege. It is a psychic fact: '4 thinks that p'
means that there is a psychic fact (involving 4) the constituents of which are
correlated with the constituents of p. These psychical constituents corre­
spond to the words of language.
I do not know what the constituents of a thought are, but I know that it
must have such constituents, which correspond to the words of language.
Again, the kind of relation of the constituents of thought and of the pic­
tured fact is irrelevant. It would be a matter of psychology to find it out
. . . The psychical constituents have the same sort of relation to reality as
words. (RUL 19.8.19; see TLP 4.1121, 5.542).
'Thinking is a kind of language' (NB 12.9.16), a thought is a proposition in
the language of thought. Although, under the pretext of anti-psychologism,
the Tractatus relegates to empirical psychology the question of what the con­
stituents of thoughts are, it incorporates the mentalist idea that it is the
mind which gives meaning to language. Representation requires an iso­
morphism between three different systems: language (propositional sign),
thought (proposition-in-thought) and reality (state of affairs) (see AWL 112;
PI §96).
What projects the psychic elements of thought onto reality? According to
one interpretation this question is misguided: unlike perceptible linguistic
propositions, thoughts are mtrinsically representational. This might be part
of their being 'logical' pictures (TLP 3), and would explain why 'the propo­
sition represents the situation, as it were, off its own bat' (NB 5.11.14). On
the other hand, it implies that the constituents of thought have precisely not
'the same sort of relation to reality as words'. It also conflicts with the idea
that meaning is conferred on signs by our conventions (TLP 3.322, 3.342,
6.53). The meanings of the 'primitive' elements of language must be
explained to us. However, since such signs are unanalysable, that is, cannot
be defined, that explanation must be by other means. The Tractatus says that
they can be explained through 'elucidations', but also that the under­
standing of these presupposes that their meanings are known (TLP 3.263,
4.026). It is therefore probable that, although the Tractatus does not mention
OSTENSIVE DEFINITION, it is acts of meaning that link a name with a particular
object, and thereby create the pictorial relation. 'By my correlating the com­
ponents of the picture with objects, it comes to represent a situation and to
be right or wrong.' T know what I mean; I mean just THIS' (NB 26.11.14,
22.6.15, see 31.5./20.6.15; TLP 2.1511). Such acts cannot be performed by
249
 MIND AND MACHINE
the empirical self, which is merely a complex of the psychic elements that
are to be correlated with objects; they must hence be acts of the 'meta­
physical' or 'willing' subject. The ineffable metaphysical subject invoked by
SOLIPSISM 'sets limits to language' by infusing words with life, a Scho­
penhauerian idea Wittgenstein later criticized (TLP 5.631, 5.641; NB 4.8./
9.11.16; PG 143-4; MS165 9-11).
In the Tractatus there is an unresolved tension between the invocation of
acts of meaning and the idea of intrinsically representational thoughts. Witt­
genstein later rejected both alternatives. The PRIVATE LANGUAGE ARGUMENT
shows that signs cannot be explained through private ostensive definitions.
He also criticized the 'old conception of the proposition' (MS 165 86), shared
by Moore, namely that propositions, unlike sentences, are immune to mis­
interpretation. Treating THOUGHTS as self-interpreting signs simply replaces a
question about indisputable capacities of sign-language (Zekhensprache) with a
mystery about the capacities of a postulated language of thought. If a word
in this language is supposed to endow spoken words with meaning, it must
itself have symbolic content. But in that case 'it would for us just be another
sign' which itself stands in need of a method of projection. This holds not
just for thought constituents, but for any 'object co-existing with the sign',
whether mental images, sense-data or Fregean senses (BB 5; see PG 40).
Associating the word 'cube' with the mental image of a cube does not deter­
mine its correct application, since that representation must itself be applied,
and could, through a suitable geometrical projection, be applied to a pyr­
amid (PI §139; see RULE-FOLLOWING).
The picture theory seems to accommodate this point, since the pictorial
relation or projection is itself an integral part of the picture, partly con­
stitutive of its being a specific picture (TLP 2.15ff.; NB 15.10.14). But this
confuses the method of projection, which cannot be part of a picture, with
the lines of projection (PI §141; PG 213-14). Even picture plus projection-
lines (the 'feelers' linking names and objects) leaves open various methods of
application, since they do not have their use laid up within them. A mental
image of two cubes connected by projection-lines can license the application
of 'cube' not only to a cube but also to a quadrangular prism. Nothing
short of the application itself determines the projected situation. But to say
that a situation is completely determined by the application of the picture
relinquishes the core of the picture theory, namely that a proposition can
picture 'off its own bat' since it is a logical form infused with content by a
method of projection. What projects signs onto reality is our using them
according to GRAMMATICAL rules (BB 4; PR 77-9, 85; PG 132; PI §§430-3).
mind and machine see HUMAN BEING
mind/body see INNER/OUTER
250
MYSTICISM
mysticism This is traditionally defined as the experience of a union with
God or the universe. Throughout his life, Wittgenstein was attracted to
unorthodox religious figures (Tolstoy, Kierkegaard, Tagore). But his only
notable treatment of mysticism is in the early work, and linked to his logico-
metaphysical system. While 'the mystical' was extremely important to Witt­
genstein, it is not the essential core of the Tractatus. Mystical themes appear
only in 1916, but then immediately dominate the Notebooks. This happened
under the influence of experiences during World War I, which led him to
read Tolstoy's Gospel in Brief, and to re-read Schopenhauer. Wittgenstein
grafted mystical themes onto a logical trunk. However, it is no coincidence
that he did so. Initially, what cannot be said but only shown are the 'logical
properties of language'. But the SAYING/SHOWING distinction invites extension
to the mystical. It promises a handle for contrasting the empirical proposi­
tions of science with not just LOGIC and metaphysics, but also 'the higher',
the realm of value - ETHICS, AESTHETICS and RELIGION. Moreover, the mystical
is the traditional archetype for something ineffable, something which 'cannot
be put into words' but 'shows itself (cp. NM 108 & TLP 6.522). Finally, the
link with mysticism keeps what is of ultimate importance, the realm of value,
safe from the encroachment of science, albeit at the price of rendering it
ineffable. At the same time, there are differences between logic and the mys­
tical. What logical propositions try to say is shown by empirical propositions.
But there are no genuine propositions which show, for example, ethical
value — although it is plausible to suppose that ethical value is shown by
peoples' actions and attitudes, as in Tolstoy's story 'The Three Hermits',
which Wittgenstein admired.
In line with, and possibly influenced by, Russell's Mysticism and Logic (ch.
X), the mystical is characterized as inexpressible (an idea Wittgenstein, but
not Russell, extended to metaphysics), yet also as involving the following:
(a) 'the problem of life', which remains untouched even if all scientific
problems have been solved (TLP 6.43ff, 6.52f.);
(b) a 'contemplation' or 'feeling' of the world sub speck aetemitate, that is,
from the outside, as a 'limited whole' (NB 7.10.16; TLP 6.45);
(c) the claim that ethics and aesthetics are based on accepting the world
(NB 20.10.16; TLP 6.42-6.43);
(d) the idea that death is unreal (TLP 643ff).
In the context of the Notebooks and the Tractatus, these familiar mystical topoi
take on a new character, (a) is related to the idea that the answer to the
problem of life is God, who is identified with 'the meaning (Sinn) of life' and
of the world (NB 11.6-/8.7.16; TLP 6.521). It has been suggested that Sinn
is here used technically, as that which is depicted by propositions (see
MEANING). But this is wrong. Firstly, that technical notion applies only to
251
 MYSTICISM
propositions, not to either life or the world. Secondly, although the Sinn that
God provides lies not in specific moral or spiritual values, it is ethical in
nature, since it consists in the 'vanishing of the problem' of life, namely as
the result of happily accepting the world as it is - (c).
Another possible link between the mystical and logical doctrines is that
the Tractatus seems to identify God with the GENERAL PROPOSITIONAL FORM,
since both are characterized as 'how things stand' (NB 1.8.16; TLP 4.5,
5.47If.). However, the general form of the proposition is 'This is how things
stand', which does not always refer to an actual fact (not all propositions are
true) but to a possible state of affairs. By contrast, God is identified with the
world understood as 'fate', something independent of our will, which sug­
gests that He is identical with how things actually are, as a matter of brute
fact. Finally, God, the meaning of life and of the world, also transcends that
world, since He 'does not reveal himself in the world' (NB 8.7.16 vs. TLP
6.41, 6.432). This could only mean that God is identical not with how the
world actually is, but with that it is.
Whether or not Wittgenstein's various claims can be made to cohere, the
last one links up with (b), the core of his mysticism. 'Not how the world is, is
the mystical, but that it is . . . The feeling of the world as a limited whole is
the mystical feeling' (TLP 6.44f.). Accordingly, the Tractatus direcdy char­
acterizes the mystical through three features:
it is the paradigm of what is 'inexpressible' and shows itself;
it is the content of an attitude, 'experience' or feeling;
it is the existence of the world.
How the world is, what the facts are, can have no value, and is part of the
problem of life, not of its solution. What is relevant to the higher is only
'that the world is'. Here there is an indisputable link between mysticism and
logic, since this is also the content of the quasi-experience presupposed by
logic: not the 'How' of the world, but its 'What': 'that something is' (TLP
5.552£). This 'experience' must concern not the truth of a contingent exis­
tential proposition, but the existence of the 'substance of the world', the
totality of simple OBJECTS. This is not to say that it is expressed by a list of
what simple objects, states of affairs or elementary propositions there actu­
ally are, which is part of the 'application of logic', not its precondition (TLP
5.55ff., 6.124). By contrast, the 'experience' at issue must be possessed by
anybody who understands propositions in their unanalysed form, not as a
conscious mental episode, but as something implicit in one's thought. What
is required is simply the knowledge that there is a totality of simple objects,
and of existing states of affairs, and that the essence or general form of pro­
positions is to say how things are. To know this is to know that the world
has limits, which might be described as knowing the world as a limited
whole.
252
MYSTICISM
The contemplation of the world sub specie aetemitatis as existing and as a
limited whole unites logic, ethics and aesthetics as 'transcendental' 'condi­
tions of the world' (NB 24.777.10./20.10.16; TLP 6.13, 6.421). But only
the last two involve (c), which cannot be milked out of Wittgenstein's meta­
physics of symbolism. Taking the world's existence for granted may be a
logical precondition of thinking, and a reflective logician might be filled with
wonder at this existence. But being content with the world, with how things
are, distinguishes the good from the evil will, and the happy from the
unhappy life.
Wittgenstein combines this idea with his peculiar version of SOLIPSISM.
Because life (the transcendental self) and world are one, the world of the
happy, that is, virtuous, man differs from that of the unhappy one (NB 29.-
30.7.16; TLP 6.43). The world 'waxes and wanes' as a whole according to
whether the transcendental self is capable of finding meaning in it, that is,
whether it accepts it in a cheerful spirit, or perceives it as a hostile place.
The other side of the solipsistic coin is that no part of the world, and no
fact, has a privileged status. This is in the first instance directed at Scho­
penhauer's idea that my own body is an embodiment of the WILL. But it also
ties in with the fact that Wittgenstein bases a Stoic moral ideal on a mys­
tical experience: T am safe, nothing can injure me, whatever happens.'Just
as the will cannot influence the world, the world cannot harm a virtuous
man. For goodness is in the eye of the beholder, in his meeting the afflic­
tions of life in a happy spirit.
Wittgenstein's solipsism is also crucial to (d). It implies that time is a
transcendental feature imposed by the metaphysical self, which is why at
death the world 'comes to an end'. At the same time Wittgenstein subscribes
to the venerable idea that eternal life belongs to those who live in the pre­
sent (TLP 6.431ff.). Happiness is attained by forsaking both fear and hope.
The way to escape the temporal character of human existence is to be con­
tent with how the world is, which is beyond the control of the human will
(TLP 6.373f.).
The early Wittgenstein manages to link traditional mystical themes with
his metaphysics of symbolism and his solipsism. Unfortunately, the construc­
tion is obscure, and there is a noticeable break between the idea that the
existence of the world is presupposed by logic - ( b ) — and the moral salva­
tion involved in accepting the world as it is — (c) and (d). This fact may
explain why the later Wittgenstein neither developed nor criticized his
earlier mysticism. By contrast, many of his readers have harped on it. For
example, it has been suggested that there are analogies between Wittgen­
stein's account of the mystical and the Zen practice of acting 'with an
empty mind'.
253
 NAMES

sense- or memory-data, the existence of which cannot be doubted. The only

logically proper names for particulars are 'this' or 'that', when used to refer
to a mental entity with which the speaker is acquainted at that moment,

N and for properties the only logically proper names are colour-terms like
'white'.
The young Wittgenstein further developed Russell's programme of LOGI­
CAL ANALYSIS. For him too, ordinary proper names are abbreviations of
descriptions. The latter are treated in accordance with a modified theory of
descriptions, the resulting existential propositions analysed as disjunctions of
elementary propositions (see GENERALITY), which are finally analysed into
n a m e s Names entered the philosophical limelight with Mill's System of
Logic (bk. I). Mill applied the label not just to proper names and common semantic atoms which are names of simple 'objects'. He did not share Rus­
nouns, but also to descriptions, abstract nouns and adjectives. At the same sell's empiricist preconceptions about what these OBJECTS must be like. He
time, he held that proper names like 'Aristotle' have a 'denotation', their was concerned primarily with showing that there must be unanalysable signs
bearer, but no 'connotation', since they do not imply an attribute. For if language is to be capable of representing reality: the PICTURE THEORY
Frege, propositions are composed of names of objects and names of con­ requires that there should be simple elements of propositions which corre­
cepts. Unlike Mill, he ascribed to ordinary proper names not just a 'mean­ spond to the indivisible elements of reality. Nevertheless, what Wittgenstein
ing', their bearer, but also a 'sense', which may differ from speaker to calls 'simple signs' or simply 'names' (TLP 3.2ff.) have to fulfil specifications
speaker: for some the sense of 'Aristode' is given by the description 'the similar to those of Russell's logically proper names.
pupil of Plato', for others by 'the teacher of Alexander the Great' ('Sense' (a) They 'go proxy for' (vertreten), 'mean' (bedeuten), 'signify' (bezeichnen) an
27). Russell took this line one step further. His logical atomism was guided object, which is their 'meaning' (Bedeutung) (TLP 2.131, 3.203, 4.0312). How­
by the 'principle of acquaintance', according to which every meaningful pro­ ever, the requirement on these objects is not epistemological (immunity from
position must consist of expressions which refer to things with which we are Cartesian doubt) but ontological: it must be impossible for them not to exist.
acquainted. Like definite descriptions ('the present King of France'), ordin­ Consequentiy, a name cannot be inserted into the argument-place of 'x
ary proper names do not fulfil this condition. The theory of descriptions exists', since the result would not be a BIPOLAR proposition.
therefore claims that the latter are really abbreviated descriptions. Definite (b) Signs which signify complexes are abbreviations (of definite descrip­
descriptions, in turn, are 'incomplete symbols', which do not refer to any­ tions, or disjunctions in the case of properties) and disappear in logical ana­
thing. Sentences of the form 'The F is G' are analysed into conjunctions of lysis (TLP 3.24). By contrast, names are 'primitive' (TLP 3.26f.). This does
three propositions: there is at least one thing which is F; there is at most not mean that they are simple qua sounds or inscriptions, but that they
one thing which is F; that thing is G. If nothing satisfies the description, cannot be further analysed or defined. They are direcdy correlated with
such sentences do not lack a truth-value, as Frege had it, but are simply objects, without the mediation of descriptions.
false (Problems ch. 5; Introduction ch. XVI). Such incomplete expressions have (c) The only descriptions of objects are propositions which say something
meaning only because they are defined through signs which cannot be about them, namely that they are combined with certain other objects, that
defined further but are direcdy linked with elements with which we are is, which state their 'external properties', but these do not tell us what an
acquainted. These are 'real' or 'logically proper names', which are ensured object is, that is, its internal properties, which specify what other objects it
against referential failure and provide the foundations of language (Logic 168, can combine with (TLP 2.023fF, 3.221). To understand a name is to grasp
194-201, 270). They stand for 'simples' (particulars, qualities and relations) its LOGICAL FORM, its combinatorial possibilities, which mirror those of the
and have the following features: (a) their meaning is an object the existence object it deputizes for. Its meaning must be explained to us, though the only
of which is not open to doubt, and to which neither existence nor non-exis­ means of explanation mentioned by the Tractatus is by 'elucidations', propo­
tence can be attributed; (b) they resist logical analysis, and are in that sense sitions which contain the sign, and hence presuppose that it is understood
'simple symbols'; and (c) to understand a logically proper name involves no
(TLP 2.0123L, 3.263, 4.026; see OSTENSIVE DEFINITION). Wittgenstein relegated
knowledge by description, only acquaintance with its meaning. From Rus­
to psychology the question how the correlation between name and object is
sell's empiricist perspective, signs which satisfy these conditions must refer to
effected (see METHOD OF PROJECTION).
There are also important differences between Russell and the Tractatus.

254 255
 NAMES
Like Russell, Wittgenstein speaks of names as having 'meaning on their
own' (TLP 3.261). But because of his CONTEXTUALISM, this means only that
they relate to reality directly, not that they have meaning in isolation, out­
side propositions. Moreover, Wittgenstein detects a lack of rigour in Russell's
account of simplicity, because Russell uses as logically proper names symbols
which we have to treat as simple because no analysis is currendy available.
For Wittgenstein, names, and the objects they stand for, are intrinsically
simple, and differ absolutely from complexes (NL 100-1; NB 26.4./21.6.15;
Logic 198, 244—6; see ELEMENTARY PROPOSITION). The price for this rigour is
Wittgenstein's refusal to provide examples of simple signs or objects. Irre­
spective of this reticence, the Tractatus is committed to rejecting Russell's
idea, condoned in the Notebooks (NB 16.6.15), that 'this' is a name. For that
implies that the meaning of a name changes on every occasion of its use,
and hence that every token of 'this' is a different name. The Tractatus, by
contrast, insists that a name is a type, the class of token-expressions which
refer to one and the same object (TLP 3.203, 3.3411; NL 102).
In this respect, Wittgenstein's later discussion of indexicals like 'this' (PI
§38; BT 523ff.; BB 109) continues his earlier work. The enterprise is a dif­
ferent one, however. Russell's claim that 'this' is the only 'genuine name' is
countered by reference not to the transcendental requirements on simple
signs, but to the actual workings of ordinary proper names. On the one
hand, there are similarities between indexicals and names: both are singular
terms, and both can occur in ostensive definitions - one can answer the
question 'What colour is your bike?' by saying either 'This colour' (pointing
to a sample) or 'Green'. On the other hand, although there are diverse
types of names (of people, places, colours, directions, numbers, etc.), an
indexical like 'this' differs from them all in at least two respects: firsdy, it
cannot be explained ostensively (partly because its referent is a function of
the context of its use); secondly, to refer, it requires accompaniment by a
deictic gesture.
Other claims of Philosophical Investigations include the Tractatus in their
target-area. The meaning of a name cannot be identified with its bearer (see
AUGUSTINIAN PICTURE OF LANGUAGE). Furthermore, there is no such thing as the
name-relation (PI §§15, 37; BB 172-3). Labels are connected with their
bearers by being attached to them, but the use of personal names and even
more-abstract names like numerals is different. The connection between a
name and its bearer is neither mysterious, nor inexorable or independent of
the way we explain and employ names (see OSTENSIVE DEFINITION). Wittgen­
stein also notices a point generally ignored by philosophers, namely the
importance proper names have for their bearer's sense of identity (GB 125—
6; MS131 141).
Investigations §79 criticizes the 'abbreviation theory' espoused by Russell
and presupposed by the Tractatus. The meaning of a proper name is not a
256
NAMES
single description which its bearer, if there is one, must uniquely satisfy.
Firstly, since different speakers associate different descriptions with a name,
this would lead to Frege's predicament, namely that the name and sentences
in which it occurs have different meanings for different speakers. Secondly,
although we may explain 'Moses' through a description, we do not treat
that description as a definition. If it turns out that such a description, for
example, 'the man who as a child was taken from the Nile by Pharaoh's
daughter', does not apply to anybody, we would not conclude that Moses
did not exist, or retract propositions about him as false, as the theory of
descriptions has it, but provide an alternative description.
Some have interpreted this as a 'cluster theory' according to which the
meaning of a name is a cluster of uniquely identifying descriptions, such
that the bearer is whatever satisfies most or a weighted proportion of them.
Investigations §79 indeed suggests a modification of the abbreviation theory in
the light of the idea of FAMILY RESEMBLANCE: the traits we use to explain
'Moses' form a loose family in which many clusters can take on the role of
being defining characteristics. But it does not commit Wittgenstein to the
underlying assumption that the meaning of names is determined by descrip­
tions, which runs counter to two other ideas of his (BT 253; PLP 71; TS211
494): (a) no definition, however complex, captures what we mean by 'Moses',
since any one could be rejected in certain circumstances - we do not use
names rigidly, in conformity with definitions laid down in advance; (b) there
are various CRITERIA for understanding proper names: giving descriptions is
not the only one, nor is it the only way of explaining names, which can also
be done through ostension or introduction - 'That is the Tower', 'I am
H.G.'
This last point also implies that 'no-meaning theories' (Mill, Kripke) are
wrong to sever names from descriptions completely. Giving a description is
one criterion, albeit defeasible, of knowing who Moses was. Nor is it clear
that the connection between name and bearer established by a baptism has
the unique role Kripke claims for it, rather than featuring among many pos­
sible explanations, as Investigations §79 suggests - 'the man who lived at that
time and place and was then called "Moses".' Finally, Kripke is wrong to
suggest that someone wmo says 'Moses was a seventeenth-century Dutch
genre-painter' must make a false statement about Moses, provided that he
has picked up the term through a communicative chain leading back to the
baptism. On the other hand, the later Wittgenstein makes an assumption
righdy questioned by no-meaning theories, namely that proper names have
a meaning which is explained by explaining who their bearer is. Only some
names have a meaning other than an etymological one. And even in those
cases, the meaning does not determine whom or what the name stands for:
'Instant Plumbing' might be the name of the slowest company in town.
257
 NATURALISM
naturalism see FORM OF LIFE; FRAMEWORK
necessity see FORM OF REPRESENTATION; LOGIC
negation see BIPOLARITY; FACT; LOGICAL CONSTANTS
nonsense For Frege, a first-level concept like 'is a planet' is a function
which maps objects onto truth-values. Any object can be the argument of
any first-level function; there are no ranges from which arguments have to
be taken. (Similarly, truth-functions admit as arguments not just propositions
but any object.) 'The number 7 is a planet' is on the same footing as 'The
sun is a planet', namely simply false. But 'is a planet is a planet' (?f(ficf) is ill-
formed, because the argument-place of the outer 'f must be filled by a satu­
rated sign, the name of an object. Although Frege introduced a hierarchy of
propositional functions, he did not operate with a distinction between sense
and nonsense ('Function' 17—21; 'Concept'; Laws I §§21-5). Russell's theory
of types, by contrast, introduced a distinction between statements which are
true or falsf and statements which are meaningless, although they may be
impeccable as regards vocabulary and syntax (Principia IT).
(1) The class of lions is a lion
is not false, as Frege had it, but 'meaningless', since it predicates of a class
what can only be predicated of individuals.
The early Wittgenstein took up this idea in a way which placed the
notion of nonsense - in the sense of 'meaningless' rather than 'obviously
false' or 'pointless' - at the centre of logic. Whether a proposition is true is
determined by how things are. LOGIC is concerned with the prior question
of what strings of signs are propositions capable of representing reality at all
(truly or falsely). He combined this with Kant's idea that philosophy is a cri­
tical activity which draws the bounds between legitimate discourse (notably,
the 'debatable sphere of science' - TLP 4.1 Iff.) and illegitimate speculation
(notably, metaphysics). The Tractatus aims
to draw a limit to thought, or rather - not to thought, but to the expres­
sion of thoughts: for in order to be able to draw a limit to thought, we
should have to find both sides of the limit thinkable (i.e. we should have
to be able to think what cannot be thought). It will therefore only be in
language that the limit can be drawn, and what lies on the other side of
the limit will simply be nonsense. (Pref.)
While Kant draws limits to knowledge, Wittgenstein draws limits to mean­
ingful discourse.
(a) Logic is concerned with thought, because it is in thought that we
258
NONSENSE
represent reality. But THOUGHTS are neither abstract nor mental entities; they
are sentences-in-use, propositional signs in their projective relation to the
world. Consequently, thoughts can be completely expressed in language,
and logic can draw limits to thought by establishing the limits of the linguis­
tic expression of thought.
(b) These limits must be drawn in language. By definition, what lies
beyond them cannot be thought, and hence - by (a) — cannot be said.
'Thought can never be of anything illogical, since, if it were, we should have
to think illogically' (TLP 3.03, 5.473, 5.61). This is impossible since logic
comprises the necessary preconditions of thought. Consequently, the limits
of thought cannot be drawn by propositions talking about both sides, but
only from the inside (TLP 4.113ff.). This is done by LOGICAL SYNTAX, the
system of'sign-rules' (Zeichenregeln) (TLP 3.32-3.34, 6.02, 6.124ff.; NM 109;
RAL 11.13) which determine whether a combination of signs is capable of
representing a possible state of affairs and hence amounts to a proposition.
(c) These rules cannot be expressed in meaningful propositions. For such
expressions would state necessary properties of symbols, and hence would
not be BIPOLAR: they would not exclude a genuine possibility, and hence
could not express what they are meant to exclude. We cannot refer to
something illogical like the class of lions being a lion by means of a mean­
ingful expression. Hence any attempt to exclude it as logically impossible is
itself nonsensical. The bounds of sense cannot be said in philosophical pro­
positions, but show themselves in the logical form of non-philosophical pro­
positions. Accordingly, the Tractatus seems committed to classifying
expressions as follows:
(i) Only the bipolar propositions of science are meaningful (TLP 4.11-
4.116, 6.53).
(ii) TAUTOLOGIES and contradictions are senseless, that is, have zero sense.
(iii) The sentences of traditional metaphysics are nonsensical. They are
based on 'misunderstandings' of logical syntax, which they violate in a
way brought out by logical analysis (TLP 3.323f., 4.003, 6.53). This
idea is retained in Philosophical Investigations §464: many metaphysical
propositions are 'latent nonsense' which GRAMMATICAL investigations
unmask by bringing out the 'patent nonsense' they imply.
(iv) The pronouncements of the Tractatus are not based on a misunder­
standing of logical syntax, but rather express insights into its workings.
In doing so, however, they try to say what can only be shown. They
are 'pseudo-propositions' which can be seen as illuminating nonsense
(TLP 4.12ff., 5.534f., 6.54f; NB 20.10.14; see SAYING/SHOWING).
The Tractatus features two accounts of nonsense. One is that the non-
sensicality of
259
 NONSENSE
(2) Socrates is identical
is a matter of deprivation, that is, due to the fact that we have failed to give
'identical' any adjectival meaning. But if we lay down such a meaning,
stipulating, for example, that 'is identical' means 'is human', then we are
dealing with two different symbols. If a proposition 'has no sense, that can
only be because we have failed to give a meaning to some of its constituents'
(TLP 5.4733, see 5.473, 6.53). Indeed, it has been suggested that according
to the Tractatus no part of (2) means what it does in a meaningful proposi­
tion. This would follow from a literal interpretation of the work's CON­
TEXTUALISM: a word (name) has meaning only in the context of a proposition
with a sense, which implies that no part of (2) has a meaning.
At the same time, the Tractatus espouses a form of compositionalism: the
sense of elementary propositions is determined by the meanings of their
constituent NAMES, that is, by what objects they stand for. Objects have a
LOGICAL FORM, which is their possibility of entering into certain combinations
with other objects. Objects, and derivatively their names, fall into different
logical categories: a point x^y in the visual field must have a colour, and
cannot have a pitch. In the case of a meaningful proposition, to grasp the
meaning and logical form of its names is to grasp the possible combination
of objects it depicts (NB 1.11.14; TLP 3.318, 4.02-4.03), while in the case of
(2) or
(3) Point xj is C-flat
it is to grasp that this combination of names does not depict a possible com­
bination of objects precisely because the constituents have incompatible
meanings. (3) amounts to what Ryle (inspired by the Tractatus) called a
'category mistake'.
Wittgenstein's later work undermines both sides of the antinomy. Con­
textualism is correct only in so far as the meaning of a word is determined
by hoW it can be USED within propositions. It does not follow that it lacks
meaning outside propositions: it is precisely the type-word on its own which
has such a use, and hence a meaning. Compositionalism regards the mean­
ing of a word as an associated entity which determines the combinatorial
possibilities of the word. Although we can talk nonsense, that is, combine
words in ways excluded by their meanings, we cannot think nonsense,
because we get stuck in trying to associate a sense with what results (see
MEANING-BODY). However, while 'The rose is red' would be nonsense if 'is'
there meant the same as in '2 x 2 is 4', the reason is not that 'is' is asso­
ciated with two different entities only one of which fits into this context.
Rather, grammar licenses the substitution of '=' in the latter, but not in the
former. This is not a consequence of 'is' having two meanings, but rather
260
NONSENSE
partly constitutive of its having a different meaning or use in these two con­
texts. Nonsense results if a combination of signs is excluded by grammar,
either through an explicit rule (e.g., 'Nothing visible can have a pitch'), or
merely by the absence of any rule for the use of an aberrant form of words,
as with (2). The fact that we could stipulate a use for (2) does not show that
it previously had a sense. We may be prevented from pursuing a path not
just by obstacles across it, but also by the path itself coming to an end (PG
53-4; PI §558, II 175-6; RPP I §§43, 246; PLP 39, 237).
Moreover, Wittgenstein criticizes compositionalism for holding that
making sense is a feature of type-sentences, and determined simply by their
form and constituents. For him (as for Ryle), it is uses of words on a parti­
cular occasion which have or lack sense. Whether an utterance makes sense,
and what sense it makes, is not determined exclusively by its linguistic form,
but depends also on the circumstances under which the utterance is made
and also on the previous communication between speaker and hearer (PI
§489, II 221; OC §§212, 229, 348-50, 433). Whether an utterance of 'This
is green' involves a category mistake depends on whether it is used to refer
to a number or an apple. A type-sentence like T see Armstrong in the
south-west corner of that room' can be used to make a perfectly intelligible
statement, but it can also be used in a nonsensical way, for example if Arm­
strong is floating in a space-ship between Jupiter and Neptune, or if I am
sitting blindfolded in my office. Conversely, saying T feel water 10 feet
down' does not make sense if the speaker has just dug a small hole into
which he reaches his arm, but would make sense if he were holding a long
probing device. The bounds of sense are not drawn once and for all by an
inexorable system, but are circumstance-relative and allow of borderline
cases (AWL 21; BB 9-10; Z §328).
Detecting nonsense in philosophy is no longer a matter of invoking a
canonical system of rules detected by LOGICAL ANALYSIS (TLP 6.53). It is
done through a critical dialogue which Wittgenstein later referred to as an
'undogmatic procedure' (WVC 183-6; see PR 54-5; BT 424-5). Persistent
misinterpretations notwithstanding, Wittgenstein refrains explicidy from
criticizing philosophical positions merely for employing words in ways that
differ from our ordinary ones (RPP I §548; RPP II §289; LPP 270). He
himself introduces technical terms where convenient. He also recognizes
that new experiences (scientific or poetic) are often expressed through
apparendy nonsensical phrases. But he would insist that this is possible
only because, in response to the new experience, a new employment of
familiar words is explained. One cannot alter the bounds of sense simply
through fiat, by uttering hitherto prohibited or vacuous forms of words.
Rather, one needs to lay down rules for the use of that form, and display
its application. Wittgenstein's ambitious claim is that it is constitutive of
metaphysical theories and questions that their employment of terms is at
261
 NONSENSE
odds with their explanations and that they use deviant rules along with the
ordinary ones. As a result, traditional philosophers cannot coherently
explain the meaning of their questions and theories. They are confronted
with a trilemma: either their novel uses of terms remain unexplained
(unintelligibility), or it is revealed that they cross language-games by using
incompatible rules (inconsistency), or their consistent employment of new
concepts simply passes by the ordinary use - including the standard use of
technical terms - and hence the concepts in terms of which the philosophi­
cal problems were phrased (PR 55-6; AWL 27; PI §191; RFM 118; LPP
7; see SCEPTICISM).
The later Wittgenstein abandons the saying/showing doctrine. Necessary
propositions other than tautologies are not pseudo-propositions. Yet, their
role is that not of empirical descriptions, but of grammatical rules (see FORM
OF REPRESENTATION). They exclude not a genuine possibility, but only a non­
sensical sign-combination. To substantiate this radical claim, Wittgenstein
relies on a bipolar principle of sense (RAL 2.7.27): the negation of a mean­
ingful proposition must also be meaningful. Yet the 'negation of an a priori
proposition' is not false (in the sense of depicting an unrealized possibility)
but nonsensical (PI §§251—2; AWL 208). Frege argued the counterposition:
although it is nonsensical to assert the negation of a logical truth, such pro­
positions are false, since their own negation is undeniably true ('Compound'
50). Wittgenstein advances three considerations for his position:
(a) One cannot think or believe a contradiction, for example that things are
not identical with themselves. This is not due to the limitations of our
powers of imagination, as Frege suggested. Nothing could even count as an
attempt to imagine such a thing. To deny the law of identity is a criterion
not of extraordinary powers of imagination, but of either a misunderstand­
ing of or a deviant use of the expression 'identical with' (PG 129-30; RFM
89-90, 95; PI §109; Laws I xvii). However, one can hold beliefs which turn
out to be contradictory, that is, cannot be spelled out coherendy, as is the
case with most philosophical theories.
(b) That a necessary falsehood cannot possibly be true means that nothing
counts as its being true. This implies, however, that one cannot specify what
the proposition asserts or means. Accordingly, the 'possibility' excluded by
necessary truths cannot be specified by the meaningful use of signs (AWL
139-43, 165-6).
(c) Our reaction to attempts to specify what it would be for a necessary
falsehood (e.g., 'This is green and yellow all over') to be true shows that we
exclude ('withdraw from circulation') a certain combination of words (PI
§§498-500; RPP II §290).
262
NONSENSE
Were we to find something which we described as green and yellow [all
over] we would immediately say this was not an excluded case. We have
not excluded any case at all, but rather the use of an expression. And
what we exclude has no semblance of sense. (AWL 63-4)
What such necessary propositions exclude is a move within a language-
game, just like 'There is no castling in draughts' (Z §134). The difference
between logically necessary and logically impossible is not akin to that
between true and false but is that between a rule of expression and a use of
words which that rule excludes as nonsensical. Grammatical propositions do
not make statements, not even, like Kant's synthetic a priori principles,
about the limits of human knowledge. They do not identify 'limits of human
understanding' which one could transcend in order to describe them.
Instead, they set the 'limit(s) of language', establish what it makes sense to
say from within (CV 15; WVC 68; BT 406-8; PI §119; BB 65). Beyond
these bounds lie not unknowable things in themselves, but only nonsense.
One might resist this conclusion by arguing that the falsehood of a neces­
sary proposition is after all conceivable or imaginable. Thus, it has been
argued that fairy-tales or Escher's drawings portray a logical impossibility.
Wittgenstein anticipated this move. There is a use of T can't imagine...'
which is an alternative way of asserting logical impossibility (Z §253; PI
§§395-7). But the bounds of sense are not determined by the scope of our
imagination (as Hume maintained - A Treatise of Human Nature I.ii.2). That
one can conjure up images in conjunction with a form of words is neither
necessary nor sufficient for its making sense (LWL 94; PG 128—9; PI §512;
Z §§247-51, 272-5; MSI 16, 65-6). To establish whether an expression
(tale/drawing) makes sense, one must investigate how it is constructed
(Escher's drawings violate rules of pictorial representation) and whether it
has an application in the language-game.
Wittgenstein also denies that there is a logical difference between gibber­
ish like 'Ab sur ah' and philosophically relevant nonsense like 'No one can
have my thought.' He admits that 'the word "nonsense" is used to exclude
different things for different reasons', but insists that 'it cannot be the case
that an expression is excluded and yet not quite excluded — excluded
because it stands for the impossible, and not quite excluded because in
excluding it we have to think the impossible.' The only difference between
ordinary and philosophical nonsense is that between patent nonsense which
causes no confusion since we recognize it immediately by the 'jingle of
words', and latent nonsense, 'where operations are required to enable us to
recognize it as nonsense' (AWL 64; PI §§464, 524; LWL 98). There is no
halfway house between sense and nonsense. This conflicts with Chomsky's
suggestion that, for example, 'Colourless green ideas sleep furiously' is syn­
tactically well-fbrmed but 'semantically anomalous'. Wittgenstein would
263
 NUMBERS
reject this position, and more generally the semiotic triad of syntax, seman­
tics and pragmatics: semantic anomalies are not propositions, since they
cannot be used to perform a move in the language-game; their lacking sense
is precisely a matter of what semioticians would regard as mere pragmatics,
namely the absence of an established use.
n u m b e r s Numbers play a crucial role in Wittgenstein's abiding rejection
of logicism, the reduction of mathematics to logic. Since all other branches
of mathematics can be build up from arithmetic, the logicist project boils
down to defining the concept of a natural number in set-theoretic terms,
and to deducing the principle of mathematical induction from logical princi­
ples. Although Frege regarded numbers as abstract objects, he effectively
defined numbers as classes of classes with the same number of members.
The number 2 is the class of pairs, the number 3 the class of trios, and so
on. This definition is not circular, since numerical equivalence between two
classes can be defined through the notion of a one-to-one correlation. Two
classes are equivalent if each member of the first can be correlated with a
different member of the other class leaving none over. The number 0 is
defined as the class of classes equivalent to the class of objects which are not
identical with themselves, that is, as a class which contains only the null-
class, {0}. The number 1 is defined as the class of classes equivalent to the
class whose only member is 0, {0}; the number 2 as the class of classes
equivalent to the class whose only members are 0 and 1, {0, 1}; etc.
Frege's ingenious procedure presupposes that classes are capable of being
members of other classes. In that case, it makes sense to ask of each class
whether or not it is a member of itself. As Russell noticed, this leads to the
paradoxical notion of the class of all classes which are not members of
themselves: if it is a member of itself, then it is not a member of itself, and
vice versa. In order to prevent the paradox, Russell introduced his theory of
types. It prohibits saying of a class what can only be said of its members,
namely that it is a member of such and such a class. Given this prohibition,
the series of natural numbers cannot be constructed in Frege's manner. The
number 1 would be of higher logical type than 0, since it has 0 as its
member, and in that case the set {0, 1} cannot be used to define the
number 2, since entities of different type cannot be members of the same
class. Russell overcame this difficulty by defining 1 as the class of all classes
equivalent to the class whose members are the members of the null-class
plus an object not a member of that class. The number 2 is defined as the
class of all classes equivalent to the class whose members are the members
of the class used to define 1 plus an object not a member of that defining
class. In this way, the natural numbers can be defined one after another,
but only if there is an infinite supply of objects. This forced Russell to intro-
264
NUMBERS
duce the axiom of infinity, according to which the number of objects in the
universe is not finite.
Russell's ad hoc use of axioms like those of infinity, of reducibility and of
multiplication or choice to prop up logicism was the first target of Wittgen­
stein's critique. These axioms are unsuitable for grounding mathematics in
logic, since they are at best contingendy true, and at worst nonsensical.
How many objects there are cannot be determined by logic (RUL 1 1 -
12.13; NB 9.10.14; TLP 5.535, 5.55, 6.1232f.; PR 167; RFM 283, 400). He
also criticized the logicist definition of the natural numbers. In later writings,
he claimed that the idea of a one-to-one correlation cannot explain the con­
cept of number. Whether there is a one-to-one correlation (e.g., between
two sets of strokes) need not simply be self-evident. And our criterion for
deciding such questions is precisely whether these sets have the same number
of elements (PG 331; PR 125-6, 281; AWL 148-9; LFM 152-68).
Wittgenstein's early criticism concerns a different point. Given the relation
between a number n and its immediate successor n+ 1, Frege and Russell
defined what it is for any number to follow n in the series of natural num­
bers, just as given y is a child of x' one can define y is a descendant of x'.
Ordinarily we would explain 'descendant' by 'a person's children, the chil­
dren of his children, the children of the children of his children, and so ori.
However, Frege and Russell felt that the 'and so on' needed to be elimi­
nated (Introduction 20-1; Foundations §§18, 79-80). To this end, they intro­
duced the notion of a hereditary or ancestral property, as one which
belongs to a person if it belongs to a person's parents. We can then define
y is a descendant of *' as y is a child of some person who has all the
hereditary properties of x\ However, while a descendant of x must have all
the properties which are hereditary in the family started by x, it is logically
possible for someone to possess all these properties without being a member
of the family. Frege and Russell sought to overcome this difficulty by claim­
ing that one of the properties hereditary in the family is precisely the prop­
erty of 'being a descendant of x\ with the consequence that anybody who
has all the hereditary properties must be a member of the family. By this
token, '« is a natural number' can be defined as 'n is identical with 0 or has
all the hereditary properties of 0', and the principle of mathematical induc­
tion turns into a logical truism: if P is a hereditary property of 0, then P
belongs to everything which has all the hereditary properties of 0.
Like Poincare, the Tractatus rejected this procedure as circular (TLP
4.1273), presumably because in defining y is a descendant of x' through the
notion of a hereditary property it treats being a descendant of x as itself a
hereditary property. Wittgenstein later suggested to Waismann (Introduction to
Mathematical Thinking ch. 8) that it is equally circular to try to establish the
principle of mathematical induction through an inductive definition of
natural number. To try to define natural numbers in such a way that the
265
 NUMBERS
principle of mathematical induction can be derived as a tautological con­
sequence is to fail to appreciate that the principle is itself a criterion for a
property's being true of all numbers, and hence partly constitutive of the
meaning of the term 'natural number'. The same reasoning informs Witt­
genstein's discussion of Skolem's inductive proof of the associative law of
addition (PG 397-424; PR ch. XTV): it is misleading to think that a proposi­
tion about all natural numbers can be proven by the principle of induction —
P(l), and P(c) entails P(c +1), therefore P holds for all numbers; for the princi­
ple defines what it is for P to hold of all natural numbers, hence it does not
provide an independent method for estabhshing such a claim.
The Tractatus treats 'number' not as a material concept which applies to
some but not all abstract objects (Frege) or classes of classes (Russell), but as
a formal concept, like that of a proposition, which is presented by a special
style of variable (TLP 6.022f.). Like 'proposition', the formal concept
'number' is expressed through a 'formal series', a series whose members are
ordered by an INTERNAL RELATION and are produced by a reiterable opera­
tion: % Q'*, Q'Q'x, fl'Q'n'*, and so on' (TLP 4.1252, 5.23ff.). We 'arrive at
numbers' as follows (TLP 6.02f). We define 'x' (the starting-point of the
+1
series) as 'Q°y, and the successor of any given member 'Q'Q*'.*' as 'Q" 'x'.
0+1 0+1+l 0 + 1 + 1+
This allows us to rewrite the series as 'Q°'x, Q ' x , fi 'x, Q 'x,
n+
etc' and to state the general form of an operation as '[Q°'x, Q"'x, Q ' ' * ] ' .
Finally, we derive the integers: 1: = 0 + 1 ; 2: = 0 + l + l; 3: = 0 + 1 + 1 + 1;
etc.
The 'general form of an integer' (TLP 6.03) (which parallels the GENERAL
PROPOSITIONAL FORM) is [0, £, £ + 1 ] . This suggests that Wittgenstein simply
provides an inductive definition of the integers which takes for granted the
notions of 0 and of the successor of a number which logicism tried to
explain. In fact, however, it is crucial to his account that numbers are not
the results of a mathematical operation (adding 1) on numerals, but fall-outs
from logical operations on propositions. 'A number is the exponent of an
operation' (TLP 6.021). Numbers correspond to stages in the construction of
molecular propositions out of elementary, ones through truth-functional
operations. This is why mathematics is a 'logical method' (TLP 6.2, 6.234).
However, unlike logicism, Wittgenstein does not regard logic as more basic
than mathematics. In the margins of Ramsey's copy of the Tractatus he
wrote 'the fundamental idea of math is the idea of calculus presented here by
the idea of operation. The beginning of logic presupposes calculation and hence
number'. Two is simply the number of times an operation must be reiter­
ated to produce an expression of the form 'Q'QV. This may appear cir­
cular: in order to define numbers it refers to the application of the operation
a certain number of times. But according to the Tractatus's SAYING/SHOWING
distinction, we need not invoke the number here: the stage of the formal
series that 'Q'Q'x' represents shows itself in the structure of that expression
266
NUMBERS
(properly analysed). In any event, it remains possible to explain the general
notion of number by reference to the idea of successive applications of an
operation. This notion in turn hinges on the idea of 'and so on'. While the
logicists unsuccessfully tried to eliminate this idea, the Tractatus makes it cen­
tral to mathematics. Both Russell and Ramsey portrayed the Tractatus's
account of mathematics as too restrictive ('Introduction'; Mathematics 17)
because it is confined to elementary numerical equations. However, its
failure to deal with transfinite cardinals is not a lacuna, but rather a
consequence of its constructivist approach. Numbers are the exponents of
operations which cannot take one beyond the finite.
Wittgenstein always rejected both the formalist and nominalist tendency
to identify numbers with numerals and the Platonist contention that numer­
als stand for abstract objects (TLP 4.241, 6.232; WVC 34, 103-5; PR 129-
30; PG 321; PI §383). Numbers are what numerals signify, but the meaning
of numerals is given not by abstract entities, but by the rules for their use.
Any sentence containing a numeral can be translated into a sentence repre­
senting the application of an operation. An equation like '2 x 2 = 4' can
2 2 4
be written as 'Q 'Q 'x = Q '*' - repeating the twofold application of an
operation twice is equivalent to its fourfold application (TLP 6.231, 6.241).
1
Arithmetic equations do not talk about numbers, they work with numbers.
A number-statement like 'There are two apples in the basket' is not about
four objects (the two apples, the basket and the number 2), but rather indi­
cates that an operation can be performed on the apples in the basket,
namely taking out one (fi'x) and taking out another (Q'Q'x).
Wittgenstein later abandoned the Tractatus's 'nebulous introduction of the
concept of number by means of the general form of an operation' (PR 131),
and treated 'number' as a FAMILY-RESEMBLANCE concept. But he retained the
idea of numbers as the product of a technique. As a result, he rejected the
notion of the actual irumite. That the series of integers is endless means not
that it refers to an abstract totality, but that the possibility of repeating the
operation ' + 1 ' is unlimited. The idea of infinity derives from the idea of an
unhmited technique of sign-construction which can be continued indef­
initely. A finite class is given by a list of its members, an infinite class is
given by a law of construction, the principle of induction (TLP 6.1232; PR
140, 160-7; PG 461; BB 95-8; RFM 138; PI §208).
Although Wittgenstein's account is constructivist, it does not amount to
finitism, let alone strict finitism. It is not driven by epistemological worries
about our capacity to apprehend infinite totalities. The proper explanation
of numbers does not take into account the feasibility of operations for
human beings. The impossibility of running through all natural numbers is
logical, not biological: there is a grammatical rule which rules out the
expression 'the greatest natural number' as nonsensical. Moreover, Wittgen­
stein does not deny that there are infinite classes, but the difference between
267
 NUMBERS

them and finite classes is not just one of size, but also a categorial one,
between an enumerable list and an unbounded operation (RFM 142; PR
148).
This idea underlies his accusation that set theory presents the difference
between finite, infinite and transfinite sets as a difference in magnitude
(WVC 228; PR 164-5, 211-22; PG 460-71; RFM 125-42). Both Dede-
kind's definition of infinity and of real numbers on the one hand, and
Cantor's proof that the rational numbers are denumerable, that is, can be
o
paired off with the natural numbers (N), while the real numbers (R) are not,
use the notion of a one-to-one correlation to establish the cardinality of
infinite sets. Given that procedure, Cantor shows that the cardinality of R is
object (Gegenstand) Wittgenstein uses this term, along with the more expli­
greater than that of N. But, as Wittgenstein points out, that procedure has a
cit 'simple object' and 'simple', for the ultimate constituents of reality (TLP
clear sense only with respect to finite classes. Cantor's diagonal method does
1.1-2.0272, 4.1272; NM 111; NB 3.9.14, 9./11.5.15; there is no evidence
not establish that R lacks an independendy defined property, having a one-
that he contrasted 'object' and 'tiling' (Ding, Sache)). Objects are essentially
to-one correlation with a given set, but extends the notion of one-to-one
simple, while 'complexes' (e.g., ordinary material objects) are combinations
correlation to infinite classes. What Cantor shows is that one can order
of simples. Objects form the 'substance of the world': since all change is the
rational numbers in a way which is precluded for real numbers. But it is
combination or separation of objects, they themselves are unchanging, and
only through a piece of concept-formation which we need not accept that
indestructible. Objects have both INTERNAL properties, their combinatorial
he reaches the conclusion that there is a hierarchy of hitherto unknown
possibilities with other objects, and external properties, being combined with
mathematical entities - the transfinite cardinals.
whatever other objects they happen to be combined with (TLP 2.01ff.). The
logical atomism of Wittgenstein and Russell seeks these elements through
LOGICAL ANALYSIS. It holds that all propositions can be revealed to be truth-
functions of atomic propositions, which in turn consist of unanalysable
NAMES. Objects are what these constituents of fully analysed propositions
stand for. They cannot be 'described', that is, defined, but only named. This
guarantees that they do not generate any necessary connections between
atomic propositions: 'The broom is in the corner' can be logically incompa­
tible with 'The mop is on the table', since mop and broom are complex,
and hence could share a common element - the handle (TIP 3.2ff.).
Russell pursued this analysis to the point where, for empiricism, the foun­
dations of language and of knowledge coincide. The existence of sense- and
memory-data is immune to doubt, which ensures that propositions about
them are immune to referential failure. Therefore the 'principle of acquain­
tance' states that we can understand propositions only if we are direcdy
aware of the simple 'individuals' they stand for. These include not just 'par­
ticulars', sense-data to which we refer through an indexical like 'this', but
also 'qualities' and 'relations'. Russell gave 'This is white' as an example of
an atomic proposition, but confessed that for all he knew 'analysis could go
on forever' (Logic 193-203, 270; Problems ch. 5).
Wittgenstein considered and rejected this possibility (NB 3.9./8.10.14).
For his logical atomism was inspired not by empiricist epistemology, but by
a quasi-Kantian theory of symbolism which explores the necessary precondi­
tions of representation. By his own admission, when Wittgenstein wrote the

268
269
 OBJECT
Tractatus he was incapable of producing an example of a simple object or an
unanalysable name. Determining the 'composition' of elementary proposi­
tions was left to future analysis. But on 'purely logical grounds' it could be
known that 'analysis must come to an end': there must be elements of reality
on the one hand and of THOUGHT and language on the other, if the latter
are to represent the former (NB 14.-17.6.15; TLP 4.221, 5.55ff.; RUL
19.8.19; AWL 11; WAM 70). That objects are postulated by a theory of
symbolism does not mean that their existence and nature are a matter of
linguistic convention (as some interpreters hold); indeed, that theory insists
that the LOGICAL FORM of names must mirror the combinatorial possibilities of
the objects they deputize for. Nor does it mean that anything whatsoever
might turn out to fulfil the role of simples. The Notebooks try hard to provide
an example of an object, and even the more agnostic Tractatus indicates the
direction in which to look. Arguably, ELEMENTARY PROPOSITIONS contain
names not just for particulars, but also for properties and relations; they
describe, for example, the colours and shapes of specks in the visual field,
and their spatial relations to other such specks. Objects are objects of
acquaintance, not sense-data but their unanalysable constituents. The closest
one gets are minima sensibilia (NB 7.5.15): particulars like spatial points, ulti­
mate perceptual qualities such as shades of colours, tones and smells, and
simple spatial relations. Unlike Russell's sense-data, they are not temporary;
they are apparent sempiternalia which are metaphysically, not just epistemi-
cally, guaranteed: red complexes and sense-data can be destroyed, the
colour red, or points in space and time, cannot; and they are incomplete:
they must combine with each other into changing combinations - i.e., facts
(see CONTEXTUALISM).
Wittgenstein's main point remains that there must be objects if repre­
sentation is to be possible. The basic idea is a regress argument going back
to Plato. Ordinary signs are explained through definitions. But the 'chain of
definitions must have an end', since defined signs signify via the signs that
serve as their definition. Consequentiy, there must be signs which relate to
objects not through definitions (descriptions) but direcdy, by naming them
(NB 9.5.15; TLP 3.26f.). Existential and universal propositions are analy-
sable into disjunctions or conjunctions of elementary propositions from
which all molecular propositions derive their sense (see GENERALITY). These
elementary propositions consist solely of names. They combine these in a
way which, given a suitable METHOD OF PROJECTION, represents a possible
combination of the objects the names stand for. No existing state of affairs
need correspond to a proposition as a whole. But unless each name were
correlated with an object, its MEANING, the proposition could not depict a
possible state of affairs. Representation requires a one-to-one correlation
between the elements of propositions and those of possible states of affairs
(TLP 4.03If, 5.123).
270
OBJECT
Thus, the PICTURE THEORY implies that names 'pin' propositions onto rea­
lity. But as Wittgenstein realized, it does not imply that their referents must
be simple or indestructible — they could be ordinary objects like books (NB
31.5.14.-15.6.15). A variety of (often implicit) considerations led him to
stronger conclusions:
(a) Complexity. The idea of a simple is 'contained' in that of a complex
which can be analysed (decomposed). There are complexes; any complex
consists of simpler parts; ergo there must be non-complex objects (NB
15.6.15; TLP 2.02ff.).
(b) The form of the world. The world has a fixed LOGICAL FORM which deter­
mines what is logically possible but is itself determined by the possibility of
things occurring in states of affairs. If all things were complex, that is,
destructible, the world's logical form, and hence what is logically possible,
could change (TLP 2.012ff.).
(c) Autonomy of sense. Whether a proposition has a sense must not depend
on the truth of another proposition stating that something or other happens
to exist, for that would make logic dependent upon contingent fact. But if
the words occurring in a proposition designated only complexes and not
(ultimately) simples, what corresponds to them might be destroyed. In that
case, they would lack meaning, since nothing would correspond to them.
Therefore, for the proposition to have a sense it would have to be true that
the corresponding complexes happen to exist, which contradicts the auton­
omy of sense (TLP 2.0211; NM 117).
To some, this argument has seemed incompatible with Wittgenstein's
rejection of truth-value gaps: a sentence with a sense must be either true or
false. A complex consists of, for example, a's standing-in-the-relation-i? to b.
A proposition ascribing a property to it - 'O(ofti)' - comes out as
'^a.m.aRb' (TLP 2.0201, 3.24; NB 15.5.15; Principles 466). If ~aRb, then
the analysandum does not lack a truth-value, but is false. It is not the sense,
but rather the truth of a proposition which depends on the existence of
complexes. But, it has a sense only if the propositions of the analysans do -
the sense of a complex proposition is a function of that of its constituents.
And these propositions are senseless unless they ultimately consist of names
for simples: if 'a' were further explained through descriptions only, its refer­
ring to something, and hence the sense of 'aRb', would depend on facts.
(d) Determinacy of sense. 'The requirement that simple signs be possible is
the requirement that sense be determinate' (TLP 3.23). Precisely why is less
determinate, but there are three possible lines of thought:
(i) Unless analysis terminates by correlating unanalysable symbols with
271
 OBJECT
simple objects, the sense of a molecular proposition, its truth-condi­
tions, is not settled in advance; we would not know, for example, how
to continue the analysis of '<S>(aRb)\ and hence what it entails (NB
18.6.15; PT 3.20101ft).
(ii) Unless the constituents of reality are sharply denned, a proposition
could not restrict reality 'to two alternatives: yes or no' (TLP 4.023).
There would not be a precise configuration of simple elements which
either verifies or falsifies it.
(iii) A proposition about a complex is indeterminate in that it can fail to
be true in more than one way: its elements can be combined in a way
which does not correspond to an existing state of affairs, or, since they
are not logically proper names, they can fail to refer. This does not
mean that such a proposition is logically deficient; for what it leaves
open is itself determinate. But this indeterminacy cannot run all the
way through to the elementary propositions into which propositions
about complexes can be analysed (TLP 5.156; NB 16.-17.6.15).
After his return to philosophy, Wittgenstein realized that he had built a
metaphysical mythology on a logical basis: in so far as the needs which
objects were meant to satisfy are genuine, they do not require necessary
existents. Argument (a) is invalid. That a complex consists of simpler parts
does not entail that there are parts which cannot be further analysed: analy­
sis might go on for ever. Equally, it is fallacious to move from the truism,
'Every complex is made up of simples' to the controversial, "There are sim­
ples of which every complex is made up.' Moreover, the distinction between
simple and complex does not have an absolute sense; one and the same
thing can be regarded as simple or complex, depending on the standards we
employ (PI §§47—8). For example, the squares of a chessboard are simple for
the purpose of the game, but might be glued together from triangular half-
squares. Indeed, for some purposes something can even be seen as com­
posed of two components greater than itself, as with the composition of
forces in mechanics.
It has been replied that not all standards of simplicity need be of equal
standing; some things might be intrinsically simple, notably the ultimate
constituents of matter. However, that would be physical simplicity, estab­
lished by experiments. What Wittgenstein is now rejecting is the idea that
there is absolute logical simplicity, required by the possibility of representa­
tion, independently of experience. In some cases (e.g., pure colours), there
may be no customary or natural standards of complexity, but this must not
be confused with the presence of absolute criteria of simplicity (PG 211; PI
§59; Z §338). And even if there were intrinsic simples, these would have to
be sempiternal only if the Tractatus were right in assuming that all change is
mere recombination of elements.
272
OBJECT
As to consideration (b), Wittgenstein's later idea of the AUTONOMY OF LAN­
GUAGE denies that there is a fixed order of logical possibilities - LOGICAL
SPACE: what is logically possible is determined not by the putative metaphysi­
cal atoms, but by linguistic rules (PI §97). By the same token, however, there
is a kernel of truth to line (c). The sense of a proposition is prior to its truth:
in order to decide whether a proposition is true, its sense must be deter­
mined (to ascribe a property F to an object a, it must be setded when some­
thing can be identified as a and what counts as being F)\ whereas to
understand its sense we need not know its truth-value, but only 'what is the
case if it is true' (TLP 4.024, 4.061f.; NB 24.10.14; PG 184-5). But to hold
that this priority requires necessary existents amounts to an argument like
Pi Necessarily: if 'aRb' makes sense, it necessarily makes sense.
P2 Necessarily: if 'aRb' makes sense, a exists.
C Necessarily: if 'aRb' makes sense, a necessarily exists.
This is valid in some systems of modal logic, but both premises are false.
There is no necessity about certain forms of words having sense. And P is 2
mistaken in assuming that a proposition has a sense only if every one of its
constituents stands for an object; most words have meaning without standing
for something, and even referring expressions like 'Excalibur' do not lose
their meaning if their referent is destroyed (PI §§39-44).
At the same time, there are expressions the meaning of which seems to be
tied to the existence of objects, namely those which can only be explained
through OSTENSIVE DEFINITION. In so far as the needs which simples were
intended to satisfy are genuine, they are fulfilled by the samples by reference
to which we explain colour-, sound- and smell-words, as well as many
measures. Like simples, these samples can be described only by specifying
their external properties, not through their internal properties. But what this
amounts to is the lexical indefinability of, for example, colour-terms. What
looked like metaphysical atoms are instruments of our form of representa­
tion. Samples are simple in that their existence is presupposed, n o t
by lan­
guage as such, but by particular language-games. Ostensively defined terms
would indeed loose their meaning if all possible samples by reference to
which they can be explained did not exist. But this does not go towards vin­
dicating (c). While the sense of an empirical proposition must not depend on
its own truth, it depends on a rule, and the possibility of explaining the rule
can depend on the truth of another proposition stating the existence of sam­
ples, or our ability to use them. Philosophical Investigations considers the reply
that at least the samples presupposed in ostensive definitions must be indes­
tructible, since otherwise we could not describe a world in which everything
destructible is destroyed, which would derogate from the independence of
grammar. But from the fact that it is possible to describe such a state of
273
 OSTENSIVE DEFINITION
affairs, it does not follow that this must be possible in that state of affairs
itself (WVC 43; AWL 120; PG 208-9; BB 31; PI §§48-57).
In consideration (d), (i) is right to claim that the logical implications of
propositions should be settled in advance of experience. However, this does
not require the existence of sempiternal simples, but only an established pat­
tern of EXPLANATION. Wittgenstein also now criticizes the assumption behind
(ii), namely that the sense of a sentence must be DETERMINATE, which itself
seems based on the image of a world of distinct facts. But even if it were
correct, it would not entail simplicity: the example of the number-line shows
that there is no incompatability between determinacy and infinite divisibility.
The same applies to (iii): many propositions can fail to be true in either of
these ways, without being the worse for it.
The later Wittgenstein ignored a second kernel of truth in the Tractatus's
account, namely the idea of non-descriptive referring. Simples were meant
to be 'what we can speak about no matter what may be the case' (PR 72). If all
referring took place through descriptions, it would require that there is one
and only one thing satisfying the description. But Donnellan has shown that
genuinely referring uses of singular terms do not depend on this condition:
one can sometimes refer to a bright young thing as 'the old fool'. However,
this requires not simples, but a shared understanding between speaker and
hearer about who is meant on this occasion.
operation see GENERAL PROPOSITIONAL FORM; LOGICAL CONSTANTS; NUMBERS;
TRUTH-TABLES
ostensive definition An ostensive definition is an explanation of the
meaning of a word as in 'This is an elephant' or 'That colour is called
"red".' It typically involves three elements: a demonstrative, 'This is...',
'That is called "..."'; a deictic gesture «S* (pointing); and a sample, the
object pointed at. The expression was first used in Johnson's Logic (1921),
but the idea itself is much older. There is a venerable position, Wittgenstein
called it the AUGUSTINIAN PICTURE OF LANGUAGE, according to which language
has a hierarchical structure. Some expressions are 'definables', that is, they
can be explained through lexical definitions. However, such definitions only
connect the definiendum with other words, the definiens. Hence it seems
that there must be 'indefinables', simple expressions which are the terminus
of lexical definitions and must be connected with objects in reality through
some kind of ostension (BB 1). This picture is prominent in Locke (An Essay
concerning Human Understanding III-4), where the objects are mental - 'simple
ideas', and a similar picture is at work in Russell, for whom logically proper
NAMES stand for individuals with which we are acquainted.
The Tractatus is equally committed to the idea that names, the simple con­
stituents of propositions which cannot be analytically defined, stand in a
274
OSTENSrVE DEFINITION
one-to-one correlation to simple objects. How, precisely, this correlation is
effected is left for psychology to determine. What is clear is that names are
related to objects direcdy, without the mediation of descriptions. Conse­
quentiy, names cannot be explained through definitions. But they can be
explained through 'elucidations . . . propositions which already contain the
primitive signs. So they can only be understood if the meanings of those
signs are already known' (TLP 3.263). This is puzzling, because the meaning
of a primitive sign will be explained through the employment of a proposi­
tion containing it, which itself can only be understood if one understands
the term explained. Perhaps the point is that we learn the meaning of
names by learning how to employ them in propositions. Accordingly, eluci­
dations would simply be elementary propositions in which the name occurs.
But this is incompatible with the idea that names, unlike propositions, need
to be explained and given meaning in the first instance (TLP 4.026-4.03).
Moreover, although Wittgenstein's early work does not mention ostensive
definition, it intimates that the METHOD OF PROJECTING names onto objects
consists of acts of ostension, of meaning this. Accordingly, in line with a sug­
gestion by Russell (Principia 91), elucidations are propositions of the form
'This is A' which explain by describing.
That, at any rate, was Wittgenstein's view after his return to philosophy.
His discussions with the Vienna Circle seem partly responsible for their inter­
est in ostensive definitions. For the logical positivists, ostensive definitions
were a means of injecting empirical content into a formal, uninterpreted cal­
culus (e.g. Papers I 219-20). Wittgenstein moved in the opposite direction.
Commenting on Waismann's attempt to summarize the Tractatus, he denied
that an ostensive definition (hinweisende Erkldrung) provides a 'connection
between language and reality', and later claimed that for this reason 'lan­
guage remains self-contained and autonomous' (WVC 209-10, 246; PG 97;
AWL 87). This is a startling claim. But Wittgenstein does not deny that we,
for the most part, talk about language-independent things; he denies only
that the latter constitute the meanings of our words, and hence that there are
semantic connections between language and world. Empirical propositions
refer to language-independent items and are verified or falsified by the way
things are. But this distinguishes them from ostensive definitions. One must
distinguish the use of a sentence of the form 'This is A' to make an empirical
statement from its employment in giving an ostensive definition (PR 54-5;
PG 88). The latter does not 'describe' the object pointed at as A, either truly
or falsely, but defines what counts as being A. An ostensive definition cannot
simultaneously be a description, just as the juxtaposition of a ruler with a rod
cannot simultaneously constitute a measurement of the rod and a calibration
of the ruler. From this perspective, the elucidations of the Tractatus are a logi­
cal mongrel between an ostensive definition of a name '4', and an empirical
proposition employing that name to depict a state of affairs.
275
 OSTENSIVE DEFINITION
Ostensive definitions have the same normative function as other types of
GRAMMATICAL explanation. They determine what counts as the correct appli­
cation of signs. For this reason, they are part of grammar (Sprachlehre), not of
the empirical application of language. More precisely, they function as sub­
stitution rules which license the substitution of a demonstrative together with
a gesture indicating a sample for the definiendum. They specify that any­
thing which is this can be characterized as being A. An ostensive definition
of red, for example, entitles one to pass from 'My bike is thisi®* colour' to
'My bike is red' (PR 78; PG 88-91, 202; BB 12, 85-90, 109). Language
remains autonomous because the samples used in ostensive definitions are
part of grammar (PI §16; PR 73). This claim does not amount to a stipula-
tive extension of the concept of language. Rather, it reminds us of the fact
that samples function as standards for the correct use of words, and thus
have a normative role analogous to that of grammatical propositions. We
explain 'Red is dusts' colour', and subsequently criticize misapplications of
the term by reference to the sample we pointed at.
The normative role of samples also lies behind Wittgenstein's claim that
the standard metre cannot be said to be (or not to be) one metre long (PI
§50). Qua sample, the object belongs to the means of representation and
cannot be described in empirical propositions. One and the same object
may function now as a sample, now as an object described as having the
defined property; but the normative and the empirical roles are mutually
exclusive inasmuch as what functions as a norm of description cannot simul­
taneously be described as falling under that norm; it might be the subject of
a subsequent measurement, but not as long as it is a canonical sample, as
was the case witivthe standard metre bar. Consequently, Kripke was right
to claim that the standard metre - this particular rod - might have had a
length other than one metre, but wrong to conclude that it therefore makes
sense to state the length of the standard metre in metres, as long as that
unit of measurement is defined by reference to that rod. What one can say
is that the rod which actually served as the standard metre might not be
used as a canonical sample, which opens up the possibility of measuring it,
but against a different standard.
Wittgenstein also uses ostensive definition to resolve a puzzle which exer­
cised the logical positivists, namely of how to explain the necessary status of
statements like 'Nothing can simultaneously be red and green all over', or
'Black is darker than white' (RFM 75-6). Such propositions cannot be
derived from explicit definitions and the laws of logic alone, that is, they are
not analytic. But neither are they synthetic a priori descriptions of the essen­
tial natures of colours. Rather, they are rules for the use of cOLOUR-words
which are part of our practice of explaining and applying these words by
reference to samples. Their necessity amounts to this: what we employ as a
sample of red we do not employ as a sample of green; and a black patch
276
OSTENSIVE DEFINITION
may serve not only to explain what 'black' means, but also, in conjunction
with a white patch, as an explanation of what 'darker than' means.
Even if one accepts that samples are part of grammar, it is natural to sup­
pose that ostensive definitions of 'indefinables' provide the ultimate explana­
tions of our words, and thereby the foundations of language. Wittgenstein
rejects this claim (BB 1; BT 256-7; see AUTONOMY OF LANGUAGE). For one
thing, whether something functions as a sample is not a matter of its essen­
tial nature, but of human choice. Samples determine the meaning of signs
only because we use them as standards of correctness. What determines the
correct use of the explanandum is not the sample, but the way we use it for
the purpose of explanation and correction. For another, not all words can
be defined ostensively, for example not 'today', 'not', 'but' or 'perhaps' (BB
77), not to mention 'relic' or 'justice'.
Moreover, ostensive definitions do not provide an inexorable grounding of
our words (PI §§27-36; LPP 427). 'An ostensive definition can be variously
interpreted in every case' (PI §28). There are several reasons for this, (a) The
deictic gesture can be misunderstood completely - thus a pupil might react
as cats do, namely look at one's hand rather than in the direction pointed
(PI §185). (b) An ostensive definition requires a stage-setting, the logico-
grammatical 'post' or category of the definiendum must be known (PI §§30-
1, 257), that is, it must be clear whether we are pointing at a colour, a
length, a shape, a number, etc. (c) We need a method of application:
acquaintance with its bearer counts for nothing, unless we master the gen­
eral use of a word, since the use of a word does not flow from the object
pointed at (see MEANING-BODY). A single connection can justify divergent pat­
terns of behaviour (PG 80). In this respect Wittgenstein has been supported
by Quine. However, for Quine ostensive definitions play only a causal role
in language acquisition by establishing dispositions to verbal behaviour. For
Wittgenstein, they have a continuing normative role in guiding our practice.
The normative trajectory of an ostensive correlation is established by a prac­
tice of correction and justification by reference to the sample.
The impossibility of providing mental equivalents for these features, espe­
cially for (c), rules out the possibility of private ostensive definitions, and
hence of a PRIVATE LANGUAGE. By a similar token, Augustine is wrong to sup­
pose that language acquisition is merely a matter of establishing a mental
association between word and object, since this presupposes the possession
of a degree of linguistic understanding, because of (b). At the same time,
there is a kernel of truth in the idea that ostensive definitions are primary.
To someone who is ignorant of either term, the explanation ' "Carnadine"
means "red"' is less useful than 'Carnadine is thisnS" colour' (PG 89-90;
RPP I §609), precisely because the latter provides him with a sample for the
application of the term. More importandy, ostension is an essential ingredi­
ent of the basic linguistic training which precedes full-blown EXPLANATIONS,
277
 OVERVIEW
whether ostensive or lexical. This is no coincidence, since we need to pre­
sent the child with paradigms to which the words apply, and often teach
words through direct exemplification ('That'sns* a banana').
This point is not necessarily inimical to Wittgenstein's account. Wittgen­
stein stressed that the fact that ostensive definitions can be misunderstood
does not mean that they are illegitimate, since this does not set them apart
from other kinds of definitions. Indeed, Wittgenstein argued that ostensive
definitions can be applied to a greater range of terms than is generally
acknowledged, including numerals (PI §§28-9). Furthermore, some expres­
sions, notably colour-terms, can be defined only ostensively (something simi­
lar holds for smells, tastes, textures and sounds). This marks a kernel of
truth in the Tractatus's myth that indestructible OBJECTS are presupposed by
language. Such terms can be explained only if suitable samples exist. One
could point at a green object and say, 'This is not red.' But this is not an
ostensive definition of either 'red' or 'not red' (PG 89-92, 136; PI 14n; BT
49-51), since such a green object cannot be used as an object of compar­
ison which would license our saying of, say, yellow or blue things that they
are not red. If all red objects suddenly turned green, the term 'red' would
not lose its meaning straightaway, as the Tractatus had it. But the technique
of applying it in, for example, 'Remember those red sunsets?' would slowly
die out. Such observations suggest that contemporary truth-conditional
theories of meaning are wrong simply to ignore ostensive definitions, which
they do in spite of the fact that their axioms relate individual constants to
objects ('«' refers to a). This means that they leave unexplained what it is for
a singular term or name to stand for an object.
overview (Ubersicht) Wittgenstein's use of this term and its cognates has
been variously rendered (e.g., as 'survey' or 'surview'), and this has cloaked
the pervasiveness and importance of the concept in his work. The need for
surveyability (Ubersichtlichkeit) of first principles was announced by Frege. But
Wittgenstein's idea that an overview of grammar provides a remedy for phi­
losophical confusion is inspired more by Hertz, who held that problems con­
cerning concepts like force are not to be answered by providing new
scientific information or definitions, but to be dissolved by a clearer under­
standing of existing information and definitions {Mechanics, Introd.). Boltz-
mann suggests, furthermore, that such dissolution is to be achieved by a
system in which the analogies or models that underlie science are presented
in a surveyable fashion (Physics 5-6, 75, 167).
Wittgenstein first introduced the term in the context of methodological
reflections on ANTHROPOLOGY (GB 130-3). He claimed that the collection of
facts concerning rituals around the world in Frazer's Golden Bough provides
not the genetic explanation of the King of Nemi ritual which Frazer
sought, but instead a different kind of illuminating synopsis of the data. He
278
OVERVIEW
contrasts the method of 'developmental hypothesis' with two other ways of
assembling the data, namely providing a 'schema', and arranging the facts
in a 'perspicuous representation'. Both are part of the 'morphological
method' of Goethe and Spengler (MSI 10 256; PLP ch. TV). Goethe's
plant-morphology uses a fictional primordial plant as an archetype by refer­
ence to which the morphology of all plants can be understood. Spengler
emulates Goethe by comparing cultural epochs to families and claiming
that cultures have archetypal life-cycles. While Wittgenstein acknowledges
Spengler's influence, he accuses him of dogmatism (CV 14-19, 26-7).
Instead of insisting that cultures must conform to his scheme, he should
have treated these 'archetypes' (Urbilder) or 'ideals' (Vorbilder) as 'objects of
comparison': they do not characterize the phenomena but determine a pos­
sible scheme for viewing them.
What is common to these thinkers is the idea that there are forms of
understanding other than the causal explanation of the deductive-nomologi-
cal sciences; and that one can shed light on a diverse multitude of phenom­
ena without discovering anything new, by arranging what is already known
in a way which clarifies the links or interconnections. Wittgenstein thought
of this methodological idea as a world-view competing with the scientistic
one. He applied it to AESTHETICS (LC 29) and MATHEMATICS. But his main use
of it is in philosophical methodology.
A main source of our failure to understand is that we do not survey
(ubersehen) the use of our words. Our grammar is lacking in perspicuity
(Ubersichtlichkeit). A perspicuous representation (ubersichtliche Darstellung) pro­
duces this understanding which consists in 'seeing connections'. Hence the
importance of finding and inventing intermediate cases. The concept of a
perspicuous representation is of fundamental significance for us. It ear­
marks our form of representation, the way we look at things. (Is this a
'Weltanschauung'?) (PI §122; see GB 133; BT 417; PR Pref; CV 7; see
also FORM OF REPRESENTATION)
This idea is heir to the Tractatus's vision of the 'correct logical point of view'
(TLP 4.1213). Both provide remedies for our 'failure to understand' gram­
mar or LOGICAL SYNTAX (the rules of language) and hence for the resulting
philosophical confusions. The Tractatus tried to achieve this aim by an ideal
notation which provides a perspicuous representation of the logical forms of
propositions without SAYING anything about them, namely through the gra­
phic means of TRUTH-TABULAR notation (TLP 4.31, 5.101, 6.1203). However,
to represent a proposition in this notation presupposes that it has been ana­
lysed. While LOGICAL ANALYSIS delves beneath the appearances of language,
the later Wittgenstein seeks a correct logical point of view by logical geo­
graphy rather than logical geology (AWL 43; LFM 44). Grammatical rules
279
 OVERVIEW
are not hidden, but lie open to view in our linguistic practice (see CALCULUS
MODEL). However, they are not perspicuous (PR 51; PI §122). Although as
competent speakers we have mastered the grammar of our language, we are
prone to misunderstand, distort or ignore certain differences between expres­
sions, or logical connections between propositions, in the course of philo­
sophical reflection.
Wittgenstein detects various roots of such philosophical confusion: (a) phe-
nomenological features both of language-use, where we associate familiar
words with specific feelings (see ASPECT-PERCEPTION) and mistakenly conclude
that they constitute the meanings of these words (PI II 174—6, 181-3, 214),
and of solitary philosophical reflection, for example, a tendency to focus
manically on a particular phenomenon to the exclusion of others (PI §§38,
593); (b) a 'craving for generality' (BB 17-18) which inclines us towards a
uniform account of FAMILY-RESEMBLANCE concepts such as 'proposition'; (c) the
emulation of science which makes us try to answer problems (e.g., the
mind/body problem) through explanatory theories instead of dissolving
them through grammatical reminders; (d) the mesmerizing influence of incli­
nations of reason, notably the 'quest for the unconditioned' (Kant), the
tendency to dig ever deeper or to look for a reality behind the phenomena
without recognizing when to stop (Z §314; RFM 102-3; RPP I §889);
(e) analogies in the surface GRAMMAR of logically distinct expressions (e.g.,
between numerals and names, or 'speaking' and 'meaning'); (f) the tendency
to project features of one language-game onto another (e.g., of discourse
about material objects onto our psychological idiom); (g) pictures embedded
in language (e.g., that things go on 'in our heads').
Wittgenstein claimed that a whole 'mythology' is laid down in our lan­
guage (GB 133; BT 433-5; PI §§422-6; O C §90; MSI 10 184). He ascribed
this idea to Paul Ernst, but the term itself occurs rather in Nietzsche, who
preceded Wittgenstein in claiming that grammatical structures may mislead
us into metaphysical illusions (The Wanderer and his Shadow §11; Beyond Good
and Evil §§16-34). Such mythologies may be harmless or even frrutful in
non-philosophical discourse, but need to be checked in philosophy, since
they obscure conceptual connections. To curb these temptations, an 'over­
view' presents a segment of grammar pertinent to a given philosophical pro­
blem in a detailed and perspicuous way (PI §122). It is an enumeration or
arrangement of grammatical rules/propositions.
Recendy, this interpretation has been challenged on the grounds that, in
spite of this professed importance, there is only one labelled instance of a
perspicuous representation in Wittgenstein's oeuvre, namely the colour-octahe­
dron, which is characterized as a 'perspicuous representation of grammar' (PR
51-2). The solution of the problem is thought to lie in the idea that perspic­
uous representations do not consist of grammatical propositions which could
take the place of grammatical explanations, but are 'second-order' descrip-
280
OVERVIEW
rions of grammatical rules. The colour-octahedron does not lay down rules
for the use of colour-words, it depicts these rules. Accordingly, there is no
dearth of perspicuous representations: any description of grammar qualifies,
and so do the simple language-games which Wittgenstein used as 'objects of
comparison'.
However, while an object of comparison (archetype, ideal) may contribute
to the provision of a perspicuous representation, it cannot constitute one by
itself, because it is supposed to 'throw light on the facts of our language by
way not only of similarities, but also of dissimilarities' (PI §§130-1, see §§2-
21; BB 77-9; CV 14, 26-7): we cannot represent the rules of chess through
describing draughts, although the latter may help us to understand the
former. The idea that perspicuous representations are on a different level
from grammatical propositions is equally untenable. 'The colour-octahedron
is grammar, since it tells us that we can speak of a reddish blue, but not of
a reddish green, etc' (PR 75; see LWL 8). This means that perspicuous
representations do consist of grammatical propositions (not of 'second-order'
descriptions).
One might nevertheless insist that for Wittgenstein there can be incom­
patible ways of articulating grammatical rules which are of equal merit.
The purpose of a perspicuous representation is not to display grammar as
it is, but to bring about a Gestalt-switch by highlighting a new aspect of
the use of our words. Perspicuous representations do not purport to be
exclusive or even correct, they aim only to remove the influence of certain
disquieting aspects of grammar, in the hope of allaying philosophical
puzzlement: 'Look at it this way..., if that doesn't calm you down, look at
it that way...' The consequence of such grammatical aspect-perception is a
philosophical relativism prominent in Waismann, who claimed that philoso­
phy should proceed through developing 'grammatical models', invented lan­
guage-games (PLP ch. IV). The idea is not to make statements about the
'reality of language', but to let these models speak for themselves. We place
them next to ordinary language and say Just look at that!' No one can
agree or disagree with this procedure, since these models do not assert any­
thing.
However, Waismann (How I see Philosophy) developed these ideas in opposi­
tion to Wittgenstein. The latter recognized that an overview establishes 'an
order' in our understanding of language which is purpose-relative (namely to
the resolution of specific problems), not 'the order' (PI §132; TS220 §107).
There are different articulations of the same grammatical rules - the colour-
octahedron could be replaced by a list of combinatorial rules ('There is no
such thing as a reddish green; there is such a thing as bluish green, etc.').
And some of them may use different objects of comparison. But none of this
implies that there is no fact of the matter as to what the grammatical rules
are or what it makes sense to say. Indeed, Wittgenstein's complaint against
281
 OVERVIEW
OVERVIEW

dogmatism is that it distorts our 'actual language' (PI §107) by projecting is that philosophy is of value because of 'the philosopher in us' (TS219 11):
onto it features of the object of comparison. This presupposes that one can the temptation to conceptual confusion is not confined to professional philo­
fail to do justice to the grammatical facts. The reply to the dogmatic 'This is sophers. But this leaves us with Ryle's famous question of what a fly would
how it has to be!' is 'Look and see' the 'application of a word as it really is' miss that never got into the fly-bottle. Here we must appreciate that philoso­
— not to make one up by concocting grammatical models (PI §§66, 112; phy should dissolve our urge to ask philosophical questions, not by whatever
MSI 11 82). The alternative to dogmatism is not relativism, but 'the quiet means (e.g., a knock on the head), but through an understanding of their
weighing of linguistic facts' (Z §447). nature and sources. A fly which never got into the botde would lack not
The notion of an overview suggests that there is a sense in which Witt- only the ability to extricate itself from similar holes, a kind of know-how,
gensteinian philosophy can be systematic. Indeed, Wittgenstein provided two but also the conceptual clarity which Wittgenstein regarded as an end in
different 'classifications of psychological concepts' (RPP I §895; RPP LI §§63, itself (PI II 206; PR Pref.; CV 7). Whether one regards a successful overview
148; Z §472). He also envisaged a 'genealogical tree' (Stammbaum) for them — as interesting in its own right, as Strawson does descriptive metaphysics, or
as for number concepts, presumably as a way of showing how, for example, merely as an aspect of philosophical critique is a matter of intellectual
the system of natural numbers can be extended into that of signed integers temperament. Even if philosophy does not contribute to human knowledge,
(RPP I §722; see PHILOSOPHICAL PSYCHOLOGY). These overviews do not aspire it contributes to human understanding.
to 'precision'. But Wittgenstein envisaged a 'complete overview of everything
which can create unclarity' (Z §§273, 464). This need not mean that there is
a 'totality' or 'complete list of rules' for our language: the notion of 'all
rules' is dubious even for a single term, since clear criteria of identity exist
only for codified rules (e.g., those of chess) (MS157a 108; TS220 92). But it
suggests that overviews of particular segments of grammar can be as com­
prehensive as one pleases.
Accordingly, there can be progress in mapping conceptual landscapes and
resolving particular problems. But this is compatible with Wittgenstein's
claim that philosophy is open-ended (Z §447; BB 44). Like the expansion of
n, philosophy can get better, without ever getting nearer to completion. The
reason is that even a global overview of grammar cannot provide a panacea
for philosophical puzzles. Firsdy, the language in which they are rooted
changes, thereby creating new problems, as happened with the development
of the new physics, of formal logic and of computers; secondly, there is not
a definite number of ways of getting confused. 'There is not a philosophical
method, though there are indeed methods, like different therapies' (PI §133).
Philosophy cannot terminate if, as Kant and Wittgenstein suggest, the fasci­
nation with philosophical problems is part of the human condition (BT 422-
4). Some passages intimate that this tendency might itself be eradicated by
cultural change (RFM 132; CV 86-9). But unlike post-modern prophets of
the demise of philosophy such as Rorty, Wittgenstein provides no clues as to
what such a change would amount to.
The final question is whether the construction of overviews constitutes a
positive aim of Wittgensteinian philosophy. Often Wittgenstein states the
aim of philosophy in purely negative terms, namely 'to show the fly the way
out of the fly-bottle' by making philosophical problems 'completely disappear'
(PI §§309, 133; see AWL 21; BT 425; CV 43). But why should one do
philosophy at all if it only gets rid of errors it itself has created? One answer

282 283

p (Wittgenstein or the interlocutor).
phenomenology see VERIFICATIONISM
Philosophical Investigations (Philosophische Untersuchungen, 1953) This is
the summa of Wittgenstein's later philosophy, just as the Tractatus was the
summa of his early work. Ever since his return to Cambridge in 1929, Witt­
genstein had tried to compose a book crystallizing his new ideas. Some 200
remarks of the printed text stem from the 'Big Typescript' of 1933. How­
ever, the work which eventually resulted in Phiksophkal Investigations started in
1936-7, after Wittgenstein had abandoned Eine Philosophische Betrachtung.
Genetically speaking, Investigations Part I falls into three parts. The first, §§1-
189, stems from the 'Early Version' (TS220). Wittgenstein offered this to
Cambridge University Press in 1938, but withdrew it within a month. Sub-
sequendy, he made several attempts to complete this trunk. The first addi­
tion (TS221, a version of Remarks on the Foundations of Mathematics Part I) is
contemporaneous with the 'Early Version' and concerns mathematics. The
second attempt took place in 1943, when Wittgenstein submitted a (now
lost) typescript to the Press. It is probable that it also included a discussion
of mathematics, since this was the main topic of Wittgenstein's manuscripts
up to 1943. This might explain why the Preface, written in 1945, still men­
tions 'the foundations of mathematics' as one of the topics discussed. In the
third attempt, the 'Intermediate Version' of 1944 (TS242), Wittgenstein
replaced the mathematical continuations by §§189-421. The final sections,
§§422-693, were added in 1945/6 (from TS228). Wittgenstein undertook
minor revisions on up to 1950, and left the book for posthumous publica­
tion.
By and large, the Investigations avoids the sibylline pronouncements of the
Tractatus. The prose is lucid and non-technical. Nevertheless, four factors
make it difficult to understand (apart from the fact that its content often
runs counter to 2,500 years of philosophizing). The first is the aphoristic
and often ironic style, which is reminiscent of Lichtenberg and Nietzsche.
Wittgenstein's remarks resonate, they provide a trajectory of thought, but
leave it to the reader to develop it. Secondly, in sharp contrast to the Trac­
tatus, the Investigations evolves around a dialogue between Wittgenstein and
284
PHILOSOPHICAL INVESTIGATIONS
an interlocutor whose confusions he tries to resolve. This dialogical structure
allows Wittgenstein to explore all the temptations and false leads presented
by a topic. Usually, the interlocutor's interventions are marked by inverted
commas. But one occasionally faces the task of determining who speaks
Thirdly, the numbered sections lack a linear structure, and there are no
formally indicated chapters. The Preface states that Wittgenstein had aban­
doned his plan of writing a more conventional, textbook-style work (this
might refer to the Blue and Brown Books and Eine Phibsophische Betrachtung) and
claims that the book travels 'over a wide field of thought criss-cross in every
direction'. Partly, this is 'connected with the very nature of the investigation',
which has to elucidate from various perspectives concepts which are them­
selves interlinked. But it is also due to a self-acknowledged failure on Wittgen­
stein's part to rein in his thoughts, which often proceed by leaps and bounds.
Fourthly, the Investigations rarely identifies its targets. As a result, some
readers have complained that Wittgenstein seems to be exorcizing views no
one has ever held. Partly, this is due to his attempt to formulate the funda­
mental assumptions and pictures which inform whole strands of philosophi­
cal thinking. But, like the lack of a linear structure, it is also due to his
idiosyncratic method of composition. The Investigations is the result of a con­
stant revision of typescripts based on first-draft manuscripts. This involved
inserting new remarks copied out from other drafts, pruning away others,
rearranging the order of remarks, curtailing particular remarks, and chan­
ging specific phrases or words. These processes had a tendency to condense
the remarks. The text became stylistically more polished, but often less intel­
ligible. Wittgenstein dropped phrases, explanations or illustrations which are
illuminating or even essential for understanding a passage, and have to be
reconstructed from the Nachlass (e.g. 46n, §§144, 373, 559).
The leitmotif which unites the various themes of Investigations Part I is lan­
guage and linguistic representation. This had already been the core of the
Tractatus, and the Preface states that the book should be seen 'by contrast
with and against the background of the Tractatus, which it criticizes for con­
taining 'grave errors'. The book starts with a quotation from Augustine, in
which Wittgenstein detects a simple picture that lies behind misconceptions
about language since Plato, but which he links in particular with Frege,
Russell and the Tractatus. Part I ends with a discussion of meaning some­
thing, which warns against regarding it as a mental activity or process.
Unlike the Tractatus, the Investigatkns pursues the connections between lin­
guistic meaning and psychological concepts, notably those of UNDERSTANDING
and THOUGHT, and volitional concepts like WILLING and INTENDING AND MEAN­
ING SOMETHING. It is mainly for this reason that it turns to issues in the philo­
sophy of mind, such as the private language argument. (These take on a life
of their own in the PHILOSOPHICAL PSYCHOLOGY after Investigations Part I.)
285
 PHILOSOPHICAL PSYCHOLOGY
In spite of its fragmentary appearance, Investigations Part I displays more
argumentative structure than is commonly assumed. One can even divide it
into 'chapters', continuous stretches of text devoted to a specific cluster of
issues:
§§1-64: the AUGUSTINIAN PICTURE OF LANGUAGE, notably in the Tractatus's
and Russell's logical atomism
§§65-88: the attack on the Tractatus's and Frege's ideal of the DETERMINACY
OF SENSE
§§89-133: the nature of PHILOSOPHY, and LOGICS quest for an ideal lan­
guage
§§134-42: the GENERAL PROPOSITIONAL FORM and the nature of TRUTH
§§143-84: linguistic understanding and the concept of reading
§§185-242: RULE-FOLLOWING and the FRAMEWORK of language
§§243-315: the PRIVATE LANGUAGE ARGUMENT
§§316-62: thought and thinking
§§363-97: IMAGINATION and mental images
§§398-411: the first-person pronoun 'i' and the nature of the self
§§412-27: CONSCIOUSNESS
§§428-65: INTENnONALITY - the harmony between language and reality
§§466-90: INDUCTION and the justification of empirical BELIEFS
§§491-546: GRAMMAR and the bounds of sense
§§547-70: IDENTITY and difference of linguistic MEANING
§§571-610: mental states and processes: expectation, belief
§§611-28: the will
§§629-60: intending
§§661-93: meaning something.
Part II of Philosophical Investigations (TS234) is part of the work on philo­
sophical psychology which preoccupied Wittgenstein after the completion of
Investigations Part I. On the basis of conversations with Wittgenstein in 1948,
the editors of the Investigations report that Wittgenstein intended to suppress a
good deal of §§491-693, and to work into its place material from Part II. But
although these parts of Part I are less polished than the preceding sections, it
is not easy to see how they could have been supplemented by material from
Part II, or how the latter could have been grafted onto Part I. Wittgenstein
never made any attempts to do so, and in many respects Part II, especially
the discussion of ASPECT-PERCEPTION, moves off in fresh directions.
philosophical psychology Themes from philosophical psychology -
intentionality, thinking, understanding - play an important role in Wittgen­
stein's later work because of their connection with linguistic meaning. How­
ever, he also had an intrinsic interest in psychology (in 1912 he conducted
286
PHILOSOPHICAL PSYCHOLOGY
experiments on the psychology of music), and after 1943 his main work was
in philosophical psychology, independendy of its connection with language.
One important cue for his discussion of the nature of psychology and of
psychological concepts in Remarks on the Phiksophy of Psychokgy was provided by
Kohler's Gestalt Psychokgy, which he read in 1947. Kohler explained the diffi­
culties of psychology by reference to its being a 'young science', which still
has to follow physics in replacing qualitative observation by quantitative
measurement. Wittgenstein rejected this diagnosis (PI II 232; RPP I §§1039,
1093). The difficulties of psychology are akin to those of set theory rather
than infant physics. They are due not to an absence of adequate instru­
mentation, a lack of quantitative concepts or a deficiency of mathematical
technique, but to conceptual confusion. Although Wittgenstein denied that
there must be a universal parallelism between the mental and the physical
(see CAUSATION; INNER/OUTER), he does not reject experimental psychology or
the study of the neurophysiological causes and prerequisites of mental phe­
nomena and abilities. His point is that the experimental methods 'pass by'
the philosophical problems, and that the latter can impede genuine advan­
ces in psychology.
Kohler preceded eliminative materialism in envisaging the possibility of
replacing our ordinary psychological statements and concepts by neuro­
physiological ones. Wittgenstein could have no qualms about empirical
psychology introducing technical terms like 'unconditioned reflex'. However,
he would insist that this will not solve the philosophical problems arising
from our ordinary psychological concepts (see METALOGIC). Moreover, for the
most part 'the concepts of psychology are just everyday concepts . . . not
concepts newly fashioned by science for its own purpose, as are the concepts
of physics and chemistry.' But in any case, ordinary language does not con­
tain a primitive theory ('folk psychology') which has been superseded by
science, as efiminative materialism has it, but only concepts like diinking, per­
ceiving, imagining (RPP II §62; Z §223). These concepts do not incorporate
a theory, since they do not predict anything and can be neither true nor
false. Instead, they are presupposed by empirical theories and define the
topics of psychology. In order to establish correlations between, for example,
perception and neurophysiological processes, it must be clear what counts as
the subject's perceiving something, which is determined by the GRAMMAR of
common-or-garden terms like 'seeing' and 'hearing'. We can modify our
psychological concepts, and have done so (e.g., by incorporating the Freu­
dian idea of the unconscious). Nevertheless, Wittgenstein complains that it is
unclear what sorts of discoveries would make possible Kohler's envisaged
replacement (MS 130 1.8.46). Moreover, even if we can establish pervasive
correlations between mental and physical phenomena, we could not aban­
don our mental concepts in favour of neurophysiological ones without ceas­
ing to be human. Although we employ these concepts inter alia to explain
287
 PHILOSOPHICAL PSYCHOLOGY
human behaviour, such explanations are not CAUSAL like those of the nomo-
logical sciences, but teleological. If we explained human behaviour as neces­
sitated by efficient causes, we would no longer treat it as intentional action,
which presupposes that the subject is guided by reasons. Moreover, unlike
neurophysiological concepts, our psychological concepts are not exclusively
or even primarily used to explain, predict and control behaviour. Their
functions are as diverse as human life (RPP II §35): we use them to express
our thoughts, emotions and attitudes, to commiserate, encourage, condemn.
These functions are essential to our life, and none of them could be served
by neurophysiological concepts.
Another inspiration for Wittgenstein's philosophical psychology was
James's encyclopaedic Principles of Psychology, which he regarded as a 'rich
source' of philosophical problems and confusions (MS 124 291; MS 165 150—
1). In contrast to Kohler's reductionism, James was a follower of Wundt's
introspective psychology. He treated introspection as an unproblematic
'looking into one's own mind', although he diverged from the Cartesian tra­
dition by admitting it to be fallible. As a result, he tried to establish the
nature of experience, thinking and the self through observing his own
'stream of thought', that is, the sequence of his mental episodes (Psychology I
185-90, 301, ch. LX). James represents what Wittgenstein called 'the old
conception' of psychology as a science which observes objects, states and
processes 'in the psychical sphere, as does physics in the physical'. But this
parallel is, he thought, misleading. Whereas the physicist observes the phe­
nomena he explains, the psychologist observes the behavioural expressions of the
mind (PI §571; TS229 §1360). According to Wittgenstein's attack on the
INNER/OUTER picture of the mind as a private realm concealed to others, this
does not mean that the subject has a more direct access to mental phenom­
ena by means of introspection. First-person present tense psychological state­
ments are typically AVOWALS, not descriptions or reports based on
observation, fallible or infallible. Moreover, inner and outer are inextricably
linked. Pace mentalism, third-person psychological propositions can be based
on what people do and say, since characteristic forms of behaviour are
CRITERIA of the mental. Pace behaviourism, we do not infer such propositions
from descriptions of mere bodily movement but describe human behaviour
ab initio in mental terms.
Even without the idea that the mind is PRIVATE, known only to the sub­
ject, the introspectionist method is awry. In line with the empiricist tradition,
it reduces all mental phenomena to mental episodes, things which cross our
mind, like feelings, sensations, images and words. In line with Kant, Witt­
genstein criticized 'the reduction of everything to sensations or images' and
the temptation to 'hypostatize feelings where there are none' (LPP 80; PI
§598). Intentional verbs like THINKING, BELIEVING or WILLING do not signify
phenomena, goings-on or states which could be detected by me, or God,
288
PHILOSOPHICAL PSYCHOLOGY
looking into my mind (PI II 217; Z §471; RPP II §§3, 31-5, 75-7, 130-3;
LW II 17-18, 74—6). The only describable 'happenings' which are relevant
to the meaning of these verbs involve overt behaviour; they do not signify
mental or neurophysiological activities, processes or states.
Wittgenstein advances two arguments in favour of this claim. One is that
inner goings-on are neither necessary nor sufficient for INTENDING AND MEAN­
ING SOMETHING. The other is that mentalists and materialists have misapplied
such categorial terms to the mental. There are mental activities (e.g., calcu­
lating in the head), events (e.g., hearing a gunshot) and processes ('under-
goings') (e.g., having mental images or impressions), and 'mental states' or
'states of consciousness' (e.g., moods - anxiety, fear, cheerfulness - or occur-
rent emotions). But there is a 'category difference' between these phenom­
ena and intentional attitudes (Z §86, see §§72-85; PI 59n, §§165, 308, 339,
572-3; RPP I §§648, 836; RPP II §§43-57, 63, 148). Intentional attitudes
are not acts or activities, since most of them are not subject to the will (one
cannot decide to, or order others to, believe or intend something) and even
those which might be, like meaning something, cannot be performed more
or less successfully (PI §§674-81; Z §§51-2). Nor are they non-voluntary
events or processes: they cannot occur, take place or go on in time, or be
slowed down, reversed or left unfinished. For example, it makes no sense to
say 'While I meant...', referring to a particular period of time. And
although 'When I said "Napoleon", I meant the victor of Austerlitz' refers to
a particular time, namely that of the original utterance, nothing need have
gone on at that time over and above the utterance (PG 103).
Although from the point of view of school-grammar, intentional verbs are
static rather than progressive, they do not signify states. States are things
one is in, but I am not currendy in a state of intending to go to London or
of believing that Napoleon was impetuous. At any given time, I believe or
intend indefinitely many things, but I am not in indefinitely many different
mental states. Wittgenstein's opponents might respond that they use 'mental
state' as a technical term, to refer to all mental phenomena. But according
to Wittgenstein, intentional attitudes differ from real 'states of consciousness'
by virtue of lacking what he calls 'genuine duration' (Z §§45-7, 81-2; RPP I
§836; RPP II §45). This means that they
(a) cannot take a course, that is, unfold in different ways;
(b) cannot be spot-checked or observed continuously;
(c) cannot be clocked by a stop-watch;
(d) are neither interrupted by a break of consciousness or a shift of atten­
tion, nor endure continuously.
This verdict can be questioned. Concerning (a) and (b), one can point out
that intentions can be more or less strong, that is, possess an observable
dimension along which they can vary without losing their identity. But
289
 PHILOSOPHICAL PSYCHOLOGY
arguably this is not an inherent property of the intention, analogous to the
persistence or intensity of a feeling, but concerns rather the manner in
which one cleaves to it. One might hold against (c) that we can measure the
time between the inception of an intention and its lapse, or realization. But
it is implausible to hold that all intentions or beliefs can be clocked in this
way. Moreover, according to (d), even those which can, lack other temporal
qualities of states. My belief that Napoleon was impetuous has not lasted
continuously for ten years, but nor has it lapsed every time I have fallen
asleep only to be resumed on waking. One can believe something inter­
mittently. Yet this is not to be interrupted in one's believing - as one's state
of anxiety may be interrupted by distractions - but to vacillate in one's
beliefs. It has been objected that (d) uses the notion of a state of conscious­
ness, and hence cannot, without circularity, demarcate such states from
intentional attitudes. But the point is simply that a state is something which
can be interrupted, whereas intentional attitudes cannot. Wittgenstein's
demarcation can be upheld only if all these different features are brought to
bear.
Accordingly, we cannot establish the essence of, for example, thinking by
watching ourselves while we think. The essence of mental phenomena, those
features which they could not fail to possess, is determined by GRAMMAR, the
rules for the correct use of mental terms. And in the case of many mental
terms these rules do not even refer to anything which crosses our mind.
Therefore philosophy should analyse mental concepts not through introspec­
tion, but by describing the use of words (PI §§314-16, 371-3, 383-4).
Later, Wittgenstein envisaged a 'genealogical tree' which would show how
different types of psychological concepts add new joints (types of moves) to
our language-games. He also tried to provide a classification of psychological
concepts or phenomena which, while not necessarily precise, would provide
a philosophically illuminating OVERVIEW (RPP I §§722, 836, 895; Z §464).
This classification suggests that psychological verbs are characterized by
first/third-person asymmetry, and can be treated as referring to mental epi­
sodes or experiences (Erkbnisse).
EXPERIENCES
undergoings emotions convictions
e.g. images e.g. sadness, e.g. belief,
joy, grief certainty, doubt
directed undirected
(attitudes)
e.g. surprise,
fright, enjoyment
impressions
290
PHILOSOPHICAL PSYCHOLOGY
Convictions like belief, certainty or doubt lack genuine duration. Under­
goings (images and impressions) have both genuine duration and intensity.
Emotions have duration, a typical mimetic expression, but lack a bodily
location (unlike sensations). They colour our thoughts - one can think sadly
or anxiously.
This classification is problematic: the categories are not clearly demar­
cated; treating experiences as the summa genera is incompatible with the
denial that convictions are goings-on; and first/third-person asymmetry does
not characterize all the terms we usually classify as mental, notably not dis­
positional ones like 'neurotic' or 'intelligent'. Wittgenstein's second attempt
(RPP II §§63, 148) is more promising.
PSYCHOLOGICAL CONCEPTS
dispositional
pain
directed undirected directed undirected
e.g. love, hate e.g. depression e.g. being e.g. anxiety
enraged by
an insult
The former category of undergoings (Erfahrungen) is divided into 'sensations'
(Sinnesempfindungen), which have genuine duration, admit of simultaneous
occurrence, degrees and qualitative mixtures, and inform us about the mate­
rial world, and 'images', which do not and are subject to the will. Emotions
are characterized by genuine duration and typical feelings. They are divided
into directed emotional dispositions (e.g., loving or hating), undirected emo­
tional dispositions (e.g., depression), undirected occurrent emotions (e.g.,
anxiety) and directed occurrent emotions (e.g., being enraged by an insult).
This classification incorporates important insights, but is still inadequate.
It fails to distinguish between sensations and perceptions, even though the
former but, following Wittgenstein, not the latter have a location in the
body. And its categories do not accommodate thinking, consciousness or
volition. Although such defects might be overcome, there is a seemingly
insuperable difficulty. Even if the domain of the psychological can be
demarcated, psychological concepts will not form a neat Porphyrian tree, as
long as they are characterized along different parameters.
291
 PHILOSOPHY
Wittgenstein neither explicitly renounced the idea of such a typology, nor
did he make further attempts to perfect it. Instead, he tried to clarify the
logical category of mental concepts by cUstinguishing them from each other
individually, through various parameters, for example according to whether
or not they have duration, an occurrent quality, phenomenal properties,
degrees, a characteristic behavioural or verbal expression, require a histor­
ical context, are subject to the will. He continued to characterize psychologi­
cal terms through the first/third-person asymmetry. But he also warns that
categorial terms like 'act', 'event', 'state' or 'process' do not provide a
sharply denned basis for philosophical psychology. 'These extremely general
terms have an extremely blurred meaning. They relate in practice to in­
numerable special cases, but that does not make them any solider; no, it
rather makes them more fluid.' Squeezing psychological concepts into cate­
gorial pigeon-holes leads only to distortion (RPP I §§257, 648; MS 167 6).
philosophy No philosopher since Kant has thought as hard about the
nature of the subject as Wittgenstein. His interest goes back to 1912, when
he gave a paper 'What is Philosophy?' In the Preface of the Tractatus he
claimed to have provided a 'definitive solution' to 'the problems of philoso­
phy'. In 1930 he maintained that his 'new method' of philosophizing was a
'kink' in the 'development of human thought' comparable to the Galilean
revolution in science. To the end of his career he insisted that what mat­
tered about his work was not its specific results but its new way of philoso­
phizing, a method or skill which would enable us to fend for ourselves (M
113-14; PI II 206; MS155 73-4; ML 17.6.41). Wittgenstein was right to
think of his methodological views as novel and radical. They run up not just
against the scientific spirit of the twentieth century (CV 6-7) but against the
whole history of philosophy. Ever since its inception, philosophy has been
regarded as akin to science in being a cognitive discipline which aspires to
provide knowledge about reality. For Platonists, philosophy is an a priori
discipline which describes not empirical reality, but a world of abstract enti­
ties, and grounds our knowledge by deducing all truths from ultimate princi­
ples. For Aristotelians, it is continuous with the special sciences because it
describes more-general features of reality — it is either the queen of the
sciences or their underlabourer, removing obstacles from their path. (Radical
empiricists (Mill, Quine) contend, furthermore, that all disciplines, including
philosophy, mathematics and logic, describe reality on the basis of empirical
evidence.)
Kant's 'Copernican Revolution' challenged this consensus. He claimed
that philosophy should be 'occupied not so much with objects as with the
mode of our knowledge of objects' (B25). While science describes reality,
philosophy is not concerned with objects of any kind, not even the abstract
entities postulated by Platonism. Instead, it reflects on the preconditions of
292
PHILOSOPHY
our knowing or experiencing the objects of the material world. Yet, in spite
of his reflective turn, Kant insisted that philosophy results in true (synthetic
a priori) propositions, those which express necessary preconditions of experi­
ence, and to that extent remains within the cognitivist tradition.
The early Wittgenstein stands in the tradition of Kant's critical philoso­
phy. Firstly, both he and Kant hold that philosophy is primarily a critical
activity which curbs the excesses of metaphysics and clarifies non-philosophi­
cal thoughts (TLP 4.112, 6.53; A l l , 735, 851). Secondly, inspired by Scho­
penhauer and Hertz, he draws a Kantian contrast between science, which
pictures or represents the world, and philosophy, which reflects on the
nature and preconditions of this representation (TLP 4.1 Iff.). This contrasts
sharply with Frege and Russell. Frege never propounded a general concep­
tion of philosophy, but his views on the nature of logic and mathematics
imply that in these two areas philosophy is a science of abstract entities.
Russell, throughout his evolutions, held fast to the 'scientific conception of
philosophy', according to which it shares the tasks of science and should
emulate its methods. While the early Wittgenstein took over Frege's separa­
tion of logic from psychology (TLP 4.1121, 6.3631, 6.423; Foundations
Introd.; Laws I Pref.) and accepted Russell's identification of philosophy and
LOGIC (TLP 4.003f.; External ch. II; Mysticism ch. 8), he argued against both
that philosophy/logic describes neither abstract objects nor the most general
features of reality, but concerns the essential preconditions of thinking about
or representing reality. He modified this Kantian idea in two respects.
Firsdy, thoughts are intrinsically linked to their linguistic expression, repre­
sentation is symbolic representation, and its preconditions are linguistic rules
- LOGICAL SYNTAX. Secondly, the bounds of sense drawn by philosophy sepa­
rate not possible knowledge from idle speculation, but meaningful from non­
sensical combinations of signs. This has drastic methodological
consequences. In his earliest discussion, Wittgenstein had claimed that philo­
sophy consists of logic (its basis) and metaphysics, and that it differs from
science in being the 'doctrine of the logical form of scientific propositions'
(NL 106). Later, he labels as 'metaphysics' exclusively the illegitimate philo­
sophy of the past. Legitimate philosophy is a 'critique of language'. 'Most of
the propositions and questions to be found in philosophical works are not
false but nonsensical' (TLP 4.003f., see Pref, 3.323-3.325, 6.51-7; NB
1.5.15, 2.12.16). They stem from the failure to understand the logic of lan­
guage, a failure which results in the asking of pseudo-questions that admit of
no answer. The task of philosophy is not to try to answer these questions
but to show that they violate the bounds of sense.
The reason for this sweeping indictment is that philosophy has always
striven to uncover necessary truths about the essential nature of the world.
But any attempt to state such necessary truths about kinds of things in the
world attributes to them formal or INTERNAL properties (e.g., that the essence
293
 PHILOSOPHY
of matter/mind is extension/thinking, that only what is extended can be
coloured, etc.). According to the SAYING/SHOWING distinction, such formal
concepts cannot significantly occur in genuine propositions, since what they
exclude is not a possibility but a piece of nonsense. The only expressible
necessary truths are 'senseless' TAUTOLOGIES; metaphysical propositions would
not be BIPOLAR and are therefore nonsensical. But what such pseudo-proposi­
tions try to say shows itself in the logico-syntactic features of non-philosophi­
cal propositions (e.g., by the fact that all names of coloured things are values
of a variable which ranges over extended things).
The propositions of the Tractatus employ formal concepts to make claims
about the essential features which language must share with reality, hence
they are themselves nonsensical (TLP 6.53). The only legitimate task of phi­
losophy is analytic and elucidatory. It neither aims at the discovery of new
truths, nor shares the piecemeal methods of the sciences. For there are no
'philosophical propositions'. Philosophy, unlike science, is not a body of doc­
trine, but an activity of clarifying non-philosophical propositions through
logical analysis (TLP 4.112). Its goal is the attainment of a 'correct logical
point of view', an understanding of what can be said (viz., empirical propo­
sitions), and of its limits. Philosophy sets 'limits to the disputable sphere of
science', 'to what cannot be thought by working outwards through what can
be thought'. Without propounding any propositions of its own, it clarifies
meaningful propositions, and demonstrates that metaphysical propositions
violate the rules of logical syntax (TLP 4.112, 4.113ff., 4.1213, 6.53).
Wittgenstein later developed this 'linguistic turn' in a different direction.
The core of his method remained the 'transition from the question of truth
to the question of meaning' (MS 106 46). The connection between philoso­
phy and language is twofold. Firsdy, there is an internal connection between
THOUGHT and its linguistic expression: philosophy is interested in language
because of its 'paramount role in human life' (BT 194-5, 413); HUMAN BEINGS
are essentially language-using animals, an idea shared by Aristotelianism and
hermeneutics. The second connection, which inspired logical positivism, is
that the a priori nature of philosophical problems and propositions is rooted
in linguistic rules: 'philosophy is the grammar of the words "must" and
"can", for that is how it shows what is a priori and what a posteriori' (CE
411). Philosophy is not a cognitive discipline, but an activity which aims at
clarity (LWL 1; AWL 225; RPP I §115). All this remains. But the ineffable
metaphysics is dropped, and the mere promise of critical analysis is replaced
by a therapeutic practice: philosophy dissolves the conceptual confusions to
which philosophical problems are alleged to owe their existence.
This picture seems to impoverish philosophy, and is generally considered
to be the weakest part of Wittgenstein's later work - slogans unsupported by
argument and belied by his own 'theory construction', which can be isolated
from the rest. Wittgenstein's methodological views must ultimately be judged
294
PHILOSOPHY
by their results. But it is important to note that they are inextricably inter­
woven with the other parts of his work, especially his conception of logical
necessity and of language, and that they arise from a coherent line of
thought:
(a) Philosophy differs in principle from the sciences because of its a priori
character.
(b) Because the a priori is to be explained by reference to linguistic rules,
it is concerned not with objects but with our way of talking about
objects according to 'grammatical rules'.
(c) These rules are not responsible to an 'essence of reality', therefore
philosophy should neither justify nor reform but only describe them.
(d) As competent speakers we are already familiar with our grammar, but
tend to ignore or distort it in philosophical reflection. Hence, describ­
ing grammar cannot lead to discoveries or theory construction; it
reminds us of how we speak, for the sake of dissolving conceptual
confusions.
(e) This conceptual clarification cannot be systematic or make progress in
the way in which science does (see METALOGIC; OVERVIEW).
(a) What links Wittgenstein's philosophizing with the metaphysical tradi­
tion is that both aim to resolve the problems which constitute the subject-
matter of philosophy (PG 193; BT 416, 431; BB 62; Z §447; PLP 5-6).
Wittgenstein suggests his 'new method' as a new way of dealing with these
problems, one which is superior because it is based on a better under­
standing of the character of the problems (LWL 1; AWL 27-8; M 113-14).
In the main, the problems concerned are those of theoretical philosophy
(logic, metaphysics, epistemology, philosophy of mind) (RW 160; M 105-6;
CV 25). Wittgenstein illustrates their peculiar nature by reference to Augus­
tine's question 'What is time?' They are a priori, and hence cannot be
solved by empirical observation or scientific experiment (AWL 3, 97, 205);
their intractable character is itself enigmatic, since they concern not the
arcane, but concepts we are familiar with in non-philosophical (everyday
and specialized) discourse; indeed, understanding these concepts is a pre­
condition of establishing new empirical facts (PI §89, see §§95, 428; BB 30-
1; BT 434-5; RPP II §289; Z §452; CV 4). One might protest with Russell
that philosophy is a proto-science, dealing with questions not yet amenable
to empirical methods (Problems 90; Logic 281). But the fact that the special
sciences developed out of philosophy does not entail that questions which
antecede experience are after all empirical. Wittgenstein argues (powerfully)
against the attempt to reduce the necessary propositions of logic, mathe­
matics and metaphysics to empirical generalizations. This is why he insists
against empiricism that philosophy is a priori (LWL 79-80). He has often
295
 PHILOSOPHY
been accused of engaging in a priori armchair science, but would respond
that it is scientistic philosophers who engage in an incoherent discipline -
empirical metaphysics.
(b) Wittgenstein's demarcation between philosophy and SCIENCE does not
express a form of irrationalism. His prohibition of theories, hypotheses and
explanations (PI §§109, 126, 496; RFM 333) bans from philosophy CAUSAL
explanations of empirical phenomena, explanations which are irrelevant to
the solution of problems that are conceptual rather than factual (Z §458;
CV 79). Socratic 'What is XV questions, in so far as they concern essential
rather than contingent features of X, arise not from ignorance about empiri­
cal reality or about a Platonist world behind appearances, but from unclar­
ity about GRAMMAR. Therefore, philosophy is not concerned with describing
or explaining reality, but with clarifying our FORM OF REPRESENTATION, which
lays down what counts as an intelligible description of reality, and, more
generally, determines what it makes sense to say.
(c) According to the Tractatus there are metaphysical truths about the
structures shared by language and reality, but they are ineffable. By con­
trast, the later Wittgenstein demythologizes metaphysics (LWL 21; MS 157(b)
4). It is constitutive of metaphysics that it confuses factual and conceptual
issues, scientific theories/hypotheses and norms of representation (Z §458;
BB 18,35). Metaphysics claims to establish true propositions about the
essence of reality. Its propositions have the form of statements of fact.
Science teaches us that no human being can run faster than 40 km/h, or
that there is no intra-mercurial planet - the metaphysician that no human
being can have the pains of another, Kant that there is no uncaused event.
According to Wittgenstein, the pronouncements of such 'descriptive' meta­
physics are grammatical rules - often distorted - in propositional disguise
(AWL 18, 65-9; WVC 67). 'Every event has a cause' is a grammatical rule
which partly determines what counts as an 'event' (see CAUSATION). By con­
trast, 'revisionary metaphysics', such as the solipsist's 'Only my present
experiences are real!', is not disguised grammar, but either nonsense or
'expressions of discontent with our grammar' (BB 55-7). Yet, grammar is
AUTONOMOUS, not responsible to a putative essence of reality. Consequentiy,
there are no metaphysical grounds for defending or reforming our grammar.
Philosophy may in no way interfere with the actual use of language; it
can in the end only describe it. For it cannot give it any foundation
either. It leaves everything as it is. (PI §124)
This dictum does not promote an intellectual quietism: Wittgenstein does
not leave philosophy as it is, but tries to reveal it as 'plain nonsense' and
'houses of cards' (PI §§118-19; BT 413, 425). Nor does it deny that lan­
guage changes (PI §18). There are non-philosophical grounds for conceptual
change (e.g., in science). The point is that it is not philosophy's business to
296
PHILOSOPHY
bring about such reform by introducing an ideal language.
(d) For the Tractatus, language must be governed by a 'super-order' of
rules, to be discovered by LOGICAL ANALYSIS. NOW Wittgenstein rejects this
CALCULUS MODEL as 'dogmatic' (PI §§81, 92, 108, 131). There are no dis­
coveries or surprises in grammar. 'If one tried to advance theses in philoso­
phy, it would never be possible to debate them, because everyone would
agree to them' (PI §128, see §599). Philosophical remarks are 'homely', 'stale
truisms' (TS213 412; MS109 212; TS220 89-90; TS219 6). Indeed, Witt­
genstein professes not to rely on 'opinions' anybody could dispute (AWL 97;
LFM 22; RFM 160; LC 72). This seems to lead to a dilemma. Either his
remarks conform to his 'no opinion'-methodology, then they cannot amount
to a genuine contribution to philosophical debate. Or they do not, then his
practice belies his stated methodological views - he would be propounding
the non-obvious thesis that there are no non-obvious philosophical theses.
Some commentators believe that Wittgenstein opted for incommensur­
ability with the philosophical tradition. According to this 'no position'-
position, his work does not contain arguments which satisfy the standards of
philosophical discourse. He is not even in the business of attacking tradi­
tional positions or correcting philosophical mistakes. His attempts to provide
a surview of grammar are therapeutic attempts to make us abandon philoso­
phical questions for the sake of intellectual tranquillity (CV 43). It is correct
that Wittgenstein did not take sides in traditional disputes, but rather tried
to undermine the assumptions common to the participants - a strategy
pioneered by Kant's 'Transcendental Dialectic' and embraced by Ramsey
(Mathematics 115-16). He also tried to 'dissolve' questions which lead to such
misguided alternatives - an idea intimated in the Preface of Moore's Princi­
pia Ethica. But in doing so, he seeks the 'right question' (see PI §§133, 189,
321; RFM 147; RPP I §600; MS130 107; WAM 27-8). And he did provide
answers to Socratic questions like 'What is understanding?', since doing so is
a prerequisite of dissolving misguided questions. What he rejects here is
merely the insistence that these questions can be answered only by analytic
definitions (see EXPLANATION). Wittgenstein was fond of comparing his philoso­
phical critique to a kind of psychotherapy (PI §§133, 254-5; BT 407-10; Z
§382). Moreover, he occasionally professed to convert us to a new point of
view (LC 27-8; CV 61). Nevertheless, his philosophical critique is an argu­
mentative rather than a medical or missionary enterprise; it should provide
arguments that are 'absolutely conclusive' (MS 161 3; BT 408, 421). Witt­
genstein does not rely on 'opinions', because he provides grammatical
reminders of how we use words outside philosophy ('It makes sense to say
"I know that she has toothache"' or 'A dog cannot be said to believe that
its master will return in a week'). Their point is to draw attention to the vio­
lation of grammar by philosophers. They are part of a dialectical critique of
sense, an 'undogmatic procedure' (WVC 183-6; see PR 54-5; PI §§89-90,
297
 PICTORIAL RELATION
127; BT 419, 424-5; LPP 45; see also NONSENSE). Wittgenstein tries to show
that his opponents use words according to conflicting rules. Some of his
remarks (e.g., 'An "inner process" stands in need of outward criteria' - PI
§580) are synoptic descriptions in which grammatical truisms are drawn
together and related to a particular philosophical problem. But even they do
not function as premises of deductive arguments. Philosophy is 'flat', without
the proofs of the deductive-nomological sciences and formal disciplines like
mathematics or logic (PI §§126, 599). Deduction establishes the con­
sequences of premises, but a dialectical critique of sense proceeds by elenchos,
not demonstration: it scrutinizes the meaning of those premises and the
inteEigibUity of the questions.
pictorial relation see METHOD OF PROJECTION
picture theory The German term Bild is ambiguous between paintings
and abstract models. 'I have inherited this concept of a picture from two
sides: first from a drawn picture, second from the model of a mathemati­
cian, which already is a general concept. For a mathematician talks of pic­
turing (Abbildung) where a painter would no longer use this expression'
(WVC 185). Hertz had claimed that science forms models of reality such
that the possible variations of the model faithfully mirror the different possi­
bilities of the physical system in question (Mechanics 1). Wittgenstein turned
Hertz's brief remarks about scientific representation into a detailed account
of the preconditions of symbolic representation in general. 'We picture facts
to ourselves' (TLP 2.1). The essence of language - the GENERAL PROPOSITIONAL
FORM - lies in depicting how things are. All meaningful propositions are
truth-functions of elementary propositions; all logical relations are due to
truth-functional composition. By accounting for elementary propositions, the
picture theory explains the basis of representation and of logic.
For this purpose, it must solve two major problems. One, which Wittgen­
stein noted (as did Frege in his latest writings), is now known as the 'crea­
tivity of language': the number of propositions is indefinite, although the
number of words is finite (NL 98; TLP 4.02, 4.027; 'Compound' 36; Post­
humous 225, 243; Correspondence 79). The other is the venerable puzzle of
intentionality, especially of explaining the possibility of falsehood. If a propo­
sition is true, it corresponds to a fact, depicts how things are in the world.
But if it is false, it remains nevertheless meaningful, although no fact corre­
sponds to it. Russell's dual-relation theory of judgement fell foul of this
puzzle: A's BELIEF that aRb cannot be a dual relation between a subject and
an object, for if it is false nothing in reality corresponds to it, which would
leave it bereft of meaning (Problems 72-3; NL 95). His multiple-relation
theory avoids the problem by holding that A is related to, 'acquainted with',
298
PICTURE THEORY
the constituents of the proposition, a, R and b, rather than the proposition
as a whole. Wittgenstein pointed out that this would allow A to judge a non­
sense, since it is no longer guaranteed that these constituents are combined
in a meaningful way. Russell tried to pluck the whole by holding that A is
acquainted not just with the constituents of the proposition, but also with a
LOGICAL FORM x<t>y, a completely general fact. Wittgenstein showed that this
conception of logical forms is inconsistent: on the one hand they are facts,
that is, complex, on the other they are objects of acquaintance, that is,
simple (NL 100-1). The first alternative creates a third-man regress: it
explains why a, R and b can combine to form certain facts (aRb, bRa) but
not others (RRb, abR) by reference to a further fact. The second simply adds
a further constituent to the proposition, without ensuring that its con­
stituents, including the additional one, are combined in a licit way.
When Wittgenstein developed the picture theory out of the ruins of Rus­
sell's theory of judgement, several points were already in place. PROPOSI­
TIONS, unlike names, are (a) essentially composite, (b) FACTS: what represents
is the fact that the proposition's components are related to each other in a
certain way, (c) BIPOLAR: they represent reality not by standing for something,
but through depicting, either truly or falsely, how things are. What
remained unsolved was the 'mystery of negation': we can say how things are
not, and the problem of the possibility of falsehood: a proposition depicts
something even if what it depicts does not obtain (NB 15.11.14). Wittgen­
stein's solution of these puzzles was that what a proposition depicts is a pos­
sibility. It does so not with the help of an additional logical form, or of
additional relations between its constituents, but simply through the fact that
its components are eombined in a certain way. The possibility of that com­
bination is guaranteed not by an additional logical form, but by the combi­
natorial possibilities of the components mirroring those of the things they
stand for.
One name is the representative of one thing, another of another thing
and they themselves are connected; in this way the whole images the
situation - like a tableau vivant... 'The connection must be possible'
means: The proposition and the components of the situation must stand
in a particular relation ... in order for a proposition to present a state of
affairs it is only necesssary for its components to represent those of the
situation and for the former to stand in a connection which is possible for
the latter. (NB 4.-5.11.14)
For a propositional sign to depict, no fact need correspond to it as a whole.
But two things are required. First, something must correspond to its ele­
ments. There must be a one-to-one correlation between these elements and
the elements of the situation it depicts. Second, it must be determined what
299
 PICTURE THEORY
relationships between the propositional elements depict what relationships
between things. If both are in place, the fact that the elements of the picture
are related to each other in a determinate way represents that the corre­
sponding things are related to each other in the same way, whether or not
they actually are. To depict falsely is to depict a non-existing combination of
existing elements. 'In a proposition a situation is, as it were, assembled by
way of experiment' (TLP 4.03 If; NB 20.-21.11.14; BB 31).
Proposition and situation must differ in some respects, and be identical in
others (NB 19.-22.10.14). On the one hand, the proposition must make
sense independendy of whether the situation is actual. On the other, they
must share a possibility which is actualized if the proposition is true, and
otherwise not. The proposition literally 'contains' that possibility (TLP 2.203,
3.02). It does not contain the content of its sense, the configuration of things
it depicts, but does contain its form, the possibility of that combination,
which is guaranteed by the logical isomorphism between the combination of
signs in the proposition and the possible combination of things in the situa­
tion (TLP 3.13, 3.34). Representation is possible through a logical isomorph­
ism, an agreement in form between what represents - whether it be
diorama, painting, musical score, proposition or thought — and what is
represented (this recalls Aristode's idea that in thought the mind and its
object, although made of different matter, take on the same form).
The pictorial nature of propositions first occurred to Wittgenstein when
he learnt about the practice of representing traffic accidents in a law court
by means of a model. We can represent a specific course of events (which
may or may not have occurred) with the aid of toy cars and dolls. To do so
we must lay down both what toy corresponds to what actual thing, and
which relations between the toys represent actual relations between objects
(e.g., their spatial relations, but not those between their weights). Thus the
Tractatus's account of propositional representation (TLP 3-4.0641) is an
application of a prior account of representation in general (TLP 2.1-2.225).
Any model or 'picture' must be composite, composed of a multiplicity of
elements going proxy for elements of the situation depicted. It must have
structure and form. The structure is the conventionally determined way in
which its elements are arranged for the purposes of depiction. The possibi­
lity of this structure is the 'pictorial form' (e.g., the three-dimensionality of a
diorama, the two-dimensionality of a painting, or the linear order of a musi­
cal score). A picture must share with what it depicts this pictorial form: it
must have the same logico-mathematical multiplicity, that is, the same
number of distinct elements, and it must be capable of combining these ele­
ments in a way mirroring the possible combinations of the objects (TLP
2.161, 4.04). Models of the same situation in different media have different
'representational forms' but the same logical form, a bare minimum of
pictorial form.
300
PICTURE THEORY
Propositional representation shares these features. A proposition is 'a
picture of reality', it describes a state of affairs by depicting it (TLP 4.016-
4.021). Elementary propositions must be composite, composed of unanaly­
sable NAMES. The 'meanings' of these names are the simple 'objects' for
which they stand, and with which they are correlated by the lines of projec­
tion. The 'sense' of the proposition is the 'state of affairs it depicts', a possi­
ble combination of objects. Only facts can represent facts, only simple
l
names can represent simple objects, and only relations (viz., that R' stands
l
in a conventionally determined relation to 'a' and b') can represent rela­
tions. A proposition consists of structure plus pictorial relation; that is, of
two relations, one between its names, and one between the latter and
objects in reality. Both of these relations are conventional: we determine not
only what names go proxy for what objects, but also what about the names
says what about the objects, that is, which relations between the names have
symbolic significance and are thereby part of the proposition's 'structure'
(TLP 3.322, 3.342, 5.473ff.). On the other hand, these conventions are
restricted by the necessary preconditions of representation, the need for a
logical isomorphism. Structure and pictorial relation are not on an equal
footing. Selecting toys as three-dimensional proxies for three-dimensional
objects is ipso facto to make their three-dimensional spatial relations the
representational form of the picture (although spatial relations between the
toys might depict, for example, relations of weight between the objects).
Moreover, the combinatorial possibilities of names must mirror those of
objects. Correlating a name with an object determines the former's combi­
natorial possibilities (TLP 3.334). Finally, once both sets of conventions are
in place, 'the proposition represents the situation, as it were, off its own bat'
(NB 5.11.14; see TLP 3.318, 4.024), irrespective of human activities. This
also solves the problem of the creativity of language: with a finite stock of
simple names and the rules of LOGICAL SYNTAX which guide their combination
in elementary propositions and also the truth-functional combination of the
resulting elementary propositions, we can form indefinitely many proposi­
tions.
The picture theory has been understood as assimilating propositions to
pictures. That suggestion cannot be rejected by insisting that propositions
are purely 'logical pictures', whose only pictorial form is logical form. For
that holds only of thoughts, not for the propositional signs which express
them (TLP 3). The latter rely on combining signs in a particular medium
(speech, writing). One might also protest that unlike a proposition, which
says that something is the case, a picture does not say anything, but can
only be used to do so. However, this ignores that the picture theory is con­
cerned with the unasserted proposition, 'the real picture in the proposition.'
Wittgenstein acknowledged that something must be added to turn such a
picture into an assertion. 'A proposition shows its sense', that is, 'how things
301
 PICTURE THEORY
stand if it is true. And it says that they do so stand' (TLP 4.022; NB
26.11.14). What is crucial to the picture theory is that even an unasserted
proposition depicts how things stand if it is true, once we have correlated its
elements with things, that is, given a method of projection. But although the
Tractatus is dominated by spatial metaphors, a proposition is a picture not in
the literal sense of relying on a spatial (or acoustic) resemblance with what it
depicts. ELEMENTARY PROPOSITIONS could not, and are not meant to, represent
relations between objects solely through the spatial arrangement of signs,
without the help of relation-names. Moreover, the 'pictorial character' (Bild-
hafhgkeit) of propositions is based not on resemblance between their elements
and those of reality, but on the 'logic of depiction' (TLP 4.01 Iff.) - the rules
of logical syntax - just as the pictorial relation between phonetic signs and
sounds consists in the existence of conventions for deriving one from the
other. A proposition is a logical picture (TLP 4.03) - albeit not a pure one.
Its pictorial nature consists in its being INTERNALLY RELATED to what it depicts;
its sense is in the proposition as the scene portrayed by a painting is in the
painting. The only respect in which LOGICAL ANALYSIS will reveal that, appear­
ances notwithstanding, propositions are pictorial, is that there is after all a
one-to-one correlation between names and objects, and more generally, that
there is a logico-mathematical isomorphism between proposition and state of
affairs (TLP 4.04). That is why Wittgenstein speaks of propositions as
'similes' or 'models', which 'construct' a world, rather than reflecting it like
a photograph (TLP 4.01, 4.023). Their pictorial nature is illustrated not just
by the literal analogy with paintings, but also by a proposition's being a
point in LOGICAL SPACE, which is determined by its constituent names just as
a point in space is determined by its coordinates (TLP 3.4ff.; NB 29.10./
1.11.14). This suggests that Wittgenstein would have resisted not only the
idea that propositions are just as realistic as paintings, but also the opposite
idea, which he helped to inspire among semioticians like Goodman, namely
that pictorial representation is just as conventional as linguistic representa­
tion.
The picture theory is a poor theory of pictures. But is it a good theory of
propositions? There has been a controversy on whether Wittgenstein later
answered this question in the negative. Some have held that the picture
theory collapses with the atomistic metaphysics with which the Tractatus
combined it. Others have insisted that there is a logico-semantical core of
the picture theory which survives into the later work. The dispute is partly
terminological, since it turns on what one includes under the labels 'picture
theory', 'logical atomism', etc. Thus, if one identifies the picture theory with
the Tractatus's overall theory of symbolism, then it collapses with the doctrine
of the general form of the proposition.
However, a pictorial conception of elementary propositions does not
essentially depend on that doctrine. This is less obvious for the doctrines of
302
PICTURE THEORY
logical atomism. Thus, it may seem that a commitment to absolutely simple,
sempiternal objects is essential to the picture theory. Now it is indeed essen­
tial to the theory that there is an object corresponding to each and every
one of a proposition's constituent names, for only then can it represent a
possibly non-existing state of affairs. However, this only yields the idea of
sempiternal and absolutely simple OBJECTS if one adds further requirements,
such as the idea that the sense of a proposition must not depend on the
contingent existence of referents for its components (autonomy of sense) or
the insistence that elementary propositions must depict definite (possible)
combinations of indestructible elements (see DETERMINACY OF SENSE).
These requirements are intimately linked to the picture theory, they pro­
vide part of its motivation, but they can nevertheless be severed from it in
principle. But even this does not hold of several ideas which Wittgenstein's
later work rightly criticizes. One is the metaphysics of facts: facts are not
composed of objects, nor are they items in the world to which TRUE proposi­
tions correspond. Moreover, the picture theory is flawed as a semantic
theory. Representation does not presuppose a one-to-one relation between
words and things. By identifying the meaning of a name with the object it
stands for, and by making a proposition's making sense dependent upon the
'meanings' of its constituent names, the picture theory subscribes to the
AUGUSTINIAN PICTURE OF LANGUAGE. Furthermore, it presupposes that a propo­
sition can represent 'off its own bat', once the structure and pictorial rela­
tion are in place. But a propositional sign cannot contain its own METHOD OF
PROJECTION. And if a proposition is identified with propositional sign plus
method of projection, depiction is guaranteed no longer by a logical form
pinned onto reality, but by our USE of the sign.
Finally, Wittgenstein attacks the core of the picture theory, the doctrine of
isomorphism. The idea that a proposition and the possible state of affairs it
depicts share a definite logical form collapses with the atomistic idea that
either of them has ultimate constituents. But without that specification, to
say that a proposition and what it depicts 'have something in common' is
merely to state that they are internally related. It is this internal relation
which Wittgenstein continues to uphold by speaking of the 'pictoriality' of
propositions (PR 57, 63-71; PG 163, 212; PI §§519-21). But this is merely
to restate the INTENTIONAL character of propositions. The explanation pro­
vided by the picture theory is rejected. The 'harmony between thought and
reality' is not a metaphysical relation between a proposition and a worldly
item (or a shadow of a worldly item - a possible state of affairs), but orche­
strated in language. It boils down to grammatical propositions like 'The pro­
position that p' = 'the proposition which is verified by the fact that p'.
Apart from the newly explained pictoriality, what remains of the picture
theory is a comparison of propositions with pictures, but with literal pic­
tures, not logical pictures (thoughts). Understanding or acting upon a propo-
303
 PRIVACY
sition is akin to understanding or acting upon a picture. The difference
between fictional and factual propositions is akin to that between genre and
historical paintings (PG 42, 163-4; WVC 185; PI §§522~3; Z §444; MS 107
155; MS 109 26-7). A genre picture tells us something, but precisely not
about how things are in reality.
privacy An essential part of the INNER/OUTER picture of the mind which
has dominated modern philosophy is the idea that mental phenomena -
ideas, sense-data, representations, experiences, etc. - are private in two
respects:
privately owned or inalienable: no one else can have my pain; other people
can at most have a pain that is similar to mine;
epistemically private: only I can know that I am in pain, since only I feel it,
others can only surmise that I am, on the basis of my behaviour.
This picture fuels scepticism about other minds: since pretence and deceit
seem always possible, one can never be sure whether other people are really
in the mental states they appear to be in according to their behaviour. It
may even lead to SOLIPSISM: if all experiential terms are defined by reference
to my inalienable experiences, it is difficult to see how there could be said to
be other subjects of experience. Wittgenstein's methodological solipsism of
the early thirties adumbrates this consequence (PR ch. VI; M 97-103). But
it avoids the idea of an ego, and opts for a no-ownership theory. Although
first-person present tense propositions about subjective experiences are
semantically basic, the first-person pronoun can be eliminated (see I/SELF). 'I
have a toothache' should read, following Lichtenberg, 'There is a tooth­
ache.' No ego or self owns these private experiences, there is only a con­
tingent causal relationship between primary experiences and a particular
body. The owner is abolished because it cannot be encountered in intro­
spection (Hume) and because of the inalienable nature of the experiences. It
is logically impossible that someone else should have what I have when I
suffer from toothache. Consequently, the T in 'I have a toothache' is redun­
dant. 'Toothache' simply means a cluster of properties which include 'being
had by me'; to ascribe it to me adds nothing. And if it is nonsense to say
that someone else has my toothache it is also nonsensical to deny it. The
Cartesian
(1) All my experiences are logically (inalienably) owned by an ego
is the product of an illusion. What we can legitimately say is empirical,
namely
(2) All my experiences are causally dependent upon a single body B.
According to Strawson, this position is incoherent, since it is forced to
304
PRIVACY
employ the notion of possession by a self or an ego which it officially rejects.
For if we drop the term 'my' from (2) we end up with something which is
simply false:
(2') All experiences are causally dependent upon a single body B.
However, for the middle Wittgenstein (2') is both true - he is a solipsist,
albeit an ego-less one - and contingent. That the pain which I call 'mine'
occurs in this body is an empirical fact, because it is conceivable that I
should suffer a pain located in someone else's body (WVC 49; BB 49-52).
The person who has a pain is the one who manifests it, and the location of
the pain is where the sufferer says it is. It is conceivable that I should locate
a toothache in someone else's mouth, for example if I flinch when his tooth
is touched, etc. But this claim falls apart in other cases. I cannot intelligibly
point to the door and say 'That's where it hurts', if the person in whose
body I locate my pain were to leave the room; or explain my limping by
saying that I have a pain in someone else's leg (PI II 222; LW II 36).
After abandoning methodological solipsism, Wittgenstein first settled for a
deflation of privacy of ownership. 'The proposition "Sensations are private"
is comparable to: "One plays patience by oneself"' (PI §248; BB 54; LW II
56). What the inner/outer picture takes to be a metaphysical truth
(3) Another person cannot have my pain
is a disguised grammatical proposition which explains the linguistic conven­
tion
(3/) My pain = the pain I have.
Whereas it makes sense to wonder whether a book belongs to me or to
someone else, it makes no sense to wonder whether the pain I feel belongs
to me or to someone else. But Wittgenstein also came to hold that there is a
sense in which (3) is confused. One argument against (3) harks back to the
no-ownership theory: 'if as a matter of logic you exclude other people's
having something, it loses its sense to say that you have it' (PI §398, see
§§253-4; MS 129 40). Another argument is that (3) is at variance with our
ascriptions of pain. Two people can have the same pain, if their pains have
the same location, intensity and phenomenal characteristics.
Nevertheless, a proponent of the inner/outer picture like Frege may reply
'Surely, another person cannot have THIS pain!' Although others can have
the same pain, that is, an exacdy similar one, they cannot have a pain
which is identical with mine ('Thought' 66—8; Foundations §27). The respective
sensations are numerically distinct, although qualitatively identical. Your
headache is in your head and mine is in my head, and by Leibniz's law,
difference in location implies numerical difference. Still, Wittgenstein insists,
even by that reasoning Siamese twins who each suffer a pain at the point of
305
 PRIVACY
juncture have the same pain. But this rejoinder is at odds with another
point he makes, one which provides a stronger reply to Frege's position. We
locate pains not by reference to spatial coordinates but by reference to parts
of the sufferer's body. If the head of one of the Siamese twins is conjoined
with the back of the other, they have different pains - the one has a head­
ache, the other a back-ache. But if you and I both have a throbbing head­
ache in the temples, we have the same pain in the same place, even though
your head is in a different place from mine.
Another response is that the insistence that my pain is mine and your
pain is yours turns the alleged owner of the pain into a custmguisbing prop­
erty of the pain, and hence makes a nonsense of the model of ownership.
Wittgenstein makes the same point against the idea of particularized quali­
ties, according to which no two objects have the same colour, since we must
distinguish the green of your chair from the green of my chair: this turns
the object into a clistinguishing property of its property, and implies that
instead of saying 'Thisns- chair is green' one should say 'This green is
1
herens ' (PR 90-1; LSD 4-5). Privacy of ownership projects onto the mental
the distinction between numerical and qualitative identity which applies only
to particulars. All we have in the mental case is the difference between A
and B's having similar pains (4's stomach-ache lacks the throbbing character
of 2Ps) and their having the same pain.
One might protest that pains do allow of a distinction between qualita­
tive and numerical identity, inasmuch as if I have a pain in my foot, and a
qualitatively indistinguishable one in my hand, I have two pains, not one.
Difference of location in the subject's body indeed implies different pains.
However, this countability of pains is restricted to each person. If two
people in a room suffer from the 'same' headache, we might say that there
are two people with headaches in the room. But it is senseless to say either
that there are two headaches in the room or that there is one. Headaches
do not have spatial locations other than their location in the body of the
sufferer.
There is one way of distinguishing qualitative and numerical identity of
pains (which Blue and Brown Books 54-5 mentions but fails to address): even if
4 and B have the 'same' pain, it is possible to destroy (anaesthetize) one of
them without destroying the other. Yet this does not derogate from the
grammatical difference between substances on the one hand, and pains and
particularized qualities on the other. Owning physical objects and owning
pains are categorially distinct: only in the former case can what is owned be
independent of the owner, and be shared between different owners.
Wittgenstein questions both aspects of epistemic privacy: 'I can know what
someone else is thinking, not what I am thinking' (PI §§246-7, II 222-4).
Third-person psychological propositions are justifiably asserted on the basis
of behavioural CRITERIA. Obviously, one can fail to manifest, or can even
306
PRIVACY
conceal, one's inner states. But these are then only ae facto unknown to
others, and can be revealed through appropriate behaviour in certain cir­
cumstances. Again, there may be individuals or communities whose emo­
tions and motives are opaque to us. Yet, we can alleviate the enigma by
learning about their biographies or way of life. Finally, there is a constitu­
tional indeterminacy of the inner, since our mental concepts do not connect
behaviour, situation and inner phenomenon in a rigid way. However, none
of these points amounts to an absolute metaphysical barrier to knowledge of
other minds (PG 82-4; LPE 314; LW II 22-31,61-73).
The other prong of Wittgenstein's attack challenges the venerable idea
that introspection provides a privileged, immediate and incorrigible knowl­
edge of our own minds. His complaint is not that introspection is after all
indirect and fallible.
(4) I know that I am in pain
is not open to ignorance or doubt. Nor does it resemble a proposition like
'N.N. is breathing' which, in normal circumstances, is simply too obvious to
have a point. For its negation is not simply false, but nonsensical. It makes
no sense to say 'I doubt whether I am in pain' or 'I have a sensation, but
don't know whether it's a pain or a tickle.' We could give sense to such
locutions, and to terms like 'unconscious pain', but only by introducing new
criteria for the use of sensation-words (BB 55).
The grammatical exclusion of grounds for the self-ascription of experience
does not show that we have an immediate and infallible access to the mind
through introspection. Equally, the unintelligibility of doubt or ignorance
does not amount to certainty or infallible knowledge - it precludes their
inteUigibility likewise. For it makes sense to say 'I know' only where it makes
sense to say 'I don't know (doubt/found out).' Like the claim that one can
only own what one could lack, this may sound like a dogmatic application
of the principle of BIPOLARITY: no knowledge without the possibility of ignor­
ance (doubt/misrecognition or misidentification), and indeed that principle
does here soldier on. However, Wittgenstein does not simply condemn (4) as
a misuse of language; he explicitly grants that it can be used. Originally he
claimed that (4) is either nonsensical, or means the same as 'I am in pain'
(BB 55; LPE 309). (4) can be used, for example, to emphasize or concede
that one is in pain, along the lines of 'I really am in pain' or 'Of course I
am in pain.' Moreover, propositions similar to (4), such as 'Only I can
know...', or 'I must know...', can be used as GRAMMATICAL propositions,
which express rules for the use of psychological terms, for example that
there is no room for doubt or that the speaker's sincere expressions have an
authoritative status, or that he can hide his sensations if he chooses (PI
§§246-8, 251-2, II 224; RPP I §§564-73).
307
 PRIVACY
But these same rules preclude (4) from expressing a genuine knowledge
claim (LPE 304—7; LSD 13, 112). There is no such thing as recognizing or
perceiving the sensation - it does not make sense to say 'From observing
myself I can tell that I am in agony.' Moreover, one does not employ cri­
teria to decide whether one has a headache or an itching sensation. Finally,
while cognitive claims exclude a possibility - doubt, ignorance or error - (4)
does not, and hence says nothing. Wittgenstein here ignores that while (4)
does not exclude as a possibility 'I do not know whether I am in pain'
(which is nonsensical, not false), it nevertheless does draw a contrast in line
with cognitive claims: '/ know that I am in pain, but she is ignorant of it.' By
contrast, 'I have my pain' does not, since both 'I do not have my pain' and
'She has my pain' are nonsensical (provided the latter is construed along the
lines of privacy of ownership). This suggests that Wittgenstein's 'bipolar' line
of argument works against privacy of ownership, but not against epistemic
privacy.
He is right, however, that there is no gap between my putative knowledge
and my simply being in pain: 'I know what I feel' = 'I feel what I feel.' ' "I
know..." may mean "I do not doubt...", but does not mean that the
words "I doubt" are senseless, that doubt is logically excluded' (PI II 221).
This does not entail that in cases in which doubt is nonsensical, it makes no
sense to speak of knowledge (one might speak of knowledge whenever some­
one is in a position to make a claim). But it shows that doing so employs
'know' in a way which lacks conceptual connections (with ways of finding
out) and contrasts (with doubt, ignorance and error) which characterize
standard uses (Z §§22, 549; LW I §51).
There remain two major lines of objection. The first denies that doubt and
ignorance are unintelligible. After all, we say things like 'While I was run­
ning, I didn't feel the pain.' But we might as well say 'While I was running, it
didn't hurt', and one could not reply 'It did, but you failed to notice.' This
equivalence is absent from ordinary epistemic claims. Again, I may not be
quite certain whether what I have is a pain or just an ache, or whether I am
really hungry. But this kind of uncertainty could not be remedied by further
evidence. Similarly, 'I don't know what I think (intend/want)' does not mean
that I think so-and-so but am ignorant of what it is, but rather that I have
not yet made up my mind. Next, one might think that it is possible to mis-
locate sensations. My dentist may show me that the tooth which hurts is the
one next to the one I thought. Someone who locates a pain in an amputated
foot seems even further off the mark. But the first case is marginal and the
second anomalous - our concept of pain-location would break down if it
became the rule. Finally, it may seem that someone who screams when a
piece of ice is put down his back falsely believes that he has a pain. But he
screamed either because the ice did pain him, or because he thought it was
going to. In neither case did he mistake one sensation for another.
308
PRIVATE LANGUAGE ARGUMENT
The second remaining major line of objection focuses on logical transfor­
mations. There are meaningful propositions which imply that (4) expresses a
genuine knowledge claim, for example
(5) I lied when I said that I was free of pain
(one can he that not-p only if one knows or believes that p), or
(6) I know that everyone in this room — including myself - is in pain.
Similarly, 'I am in pain' is a base for sentence-forming operators which
result in propositions that definitely express something I can know, such as
(7) I was in pain.
But these transformations cannot be items of knowledge, unless the base is
as well.
Wittgenstein's rejoinder to this line of objection (LPE 280, 293 4; LW II
33) is based on the idea that the possibility of such transformations does not
settle the status of (4). Granted that (5)-(7) involve knowledge claims, the
question is what that knowledge amounts to. And the answer must partly
refer back to the status of (4). Thus, Wittgenstein insists that (5), unlike lying
about one's height, does not presuppose knowledge. He could also insist that
(6) should be explicated as
(6') I am in a position to say that everyone in this room is in pain, since
I am in pain and I know that everyone else is in pain.
The kernel of truth in epistemic privacy is first-person authority: I am in
a position to say what I feel, experience, think, not because I have infallible
access to a private peep-show, but because what I say, unlike what others
say about me, is (typically) an AVOWAL, a groundless expression or manifesta­
tion of the inner.
private language argument In a wide sense, this expression refers to
the investigation of the relationship between the mental and behaviour in
Philosophical Investigations §§243-315. More narrowly, it refers to a line of
argument which discusses the idea of a 'private language' (MS 165 101-2).
Such a language is not a personal code (as the one used in some of Witt­
genstein's notebooks), nor a language which is used only in soliloquy (as the
one canvassed in §243), nor even a language spoken by only one person (as
that of the Robinson Crusoe envisaged in MS 124 221). It is not a language
which is unshared as a matter of fact, but one which is unsharable and
unteachable in principle, because its words refer to what can only be known
to the speaker, namely his immediate private experiences. §§243-55 intro­
duce the idea of a private language and show that our psychological voca-
309
 PRIVATE LANGUAGE ARGUMENT
bulary is not private in this sense, while §§256~71 argue that the very notion
is incoherent, and §§272-315 that this does not imply that the mental is
unreal.
The possibility of a private language is tacidy presupposed by the main­
stream of modern philosophy from Descartes through classical British
empiricism and Kantianism to contemporary cognitive representationalism.
It is the result of two natural assumptions. Firsdy, the meaning of words is
given by what they stand for - this is part of the AUGUSTINIAN PICTURE OF
LANGUAGE. Secondly, in the case of psychological terms, what they stand for
are phenomena in a mental theatre which is accessible only to the indivi­
dual. Sensations, experiences, thoughts are inalienable and epistemically pri­
vate (see PRIVACY). NO one else can have my pain, or know what I have
when I am in pain - this is the INNER/OUTER picture of the mind. It follows
immediately that no one else can know what I mean by 'pain'. Moreover, if
ideas, impressions or intuitions provide not just the evidence for all our
beliefs, but also the content of all our words - a view shared by representa-
tionalists and idealists, rationalists, empiricists and Kantians - our whole lan­
guage is private in this sense.
The idea that meanings are private experiences raises Locke's spectre of
the inverted spectrum (An Essay concerning Human Understanding LI.32.15): for
all we know, what I mean by 'red' is what you mean by 'green'. The first to
accept this conclusion was Russell (Logic 195). So possessed was he by the
idea that the meanings of our words must be sense-data with which we are
acquainted, that he declared it to be a precondition of intersubjective
UNDERSTANDING that no two people mean the same by their words. The
Tractatus moved along similar lines. Although the OBJECTS which are the
meanings of simple NAMES are not sense-data, they are objects of acquain­
tance - shades of colours, points in the visual field. In his VERIFICATIONIST
phase, Wittgenstein, in step with Carnap and Schlick, held that there is a
primary 'phenomenological' language which refers to immediate experi­
ences. Between 1932 and 1935, he first relinquished the idea of a primary
language, and next attacked idealism and SOLIPSISM. The notion of a private
language first appears in lectures of 1935-6 (LPE; LSD); the argument
against it is developed in manuscripts of 1937-9 and completed in 1944-5.
The final version, polished but condensed, is Investigations §§243-315. This
has been denied by some adherents of a community-view of RULE-FOLLOWING,
who claim that the 'real' private language argument is completed by §202,
which states that 'it is not possible to follow a rule "privately": otherwise
thinking one was obeying a rule would be the same thing as obeying it.' On
this view, §§243-315 merely defend the idea that meaningful discourse
requires an actual community of speakers against a potential counter-exam­
ple, sensation-terms. But in the original drafts (MS 180a 68-72; MS 129 116-
17), §202 follows and presupposes §§243-315. Moreover, the discussion of a
310
PRIVATE LANGUAGE ARGUMENT
private language in §§243-315 is not concerned with soliloquists who do not
communicate their sensations. Finally, it does not just apply a lesson about
rule-following to sensations, but attempts to dispel general misconceptions
about the mind (mental states and processes) and its relationship to beha­
viour. Although its main focus is on sensations (Empfindungen), and in parti­
cular on pain, it is equally concerned with experience, notably with visual
experiences (PI §§273-80, 290, 305-6, 312).
On the other hand, the private language argument does presuppose the
antecedent discussion of rule-following. A discussion of the coherence of the
notion of a private language presupposes a conception of language, and
Wittgenstein regards language as an activity guided by GRAMMATICAL rules.
But he did not reach the conclusion that a private language is impossible
simply by defining language as a means of communication (see FAMILY
RESEMBLANCE), or by applying a previously established 'community-view' of
rules. The connection is rather that rules are standards of correctness, and
that for a sign like 'pain' to be the name of a sensation, and not just a
squiggle or noise, it must be determined how it is to be used (LPE 291). In
a putative private language no such standard of correctness could be set up
or employed, and hence its signs would be meaningless. A language which is
in principle unintelligible to anyone but its speaker is not just (trivially)
unsuited to communication, it is unintelligible even to the private linguist
himself. The private linguist claims in our public language that he is using a
sign 'S' as part of a language, that is, according to rules, albeit rules only he
understands (PI §§261, 270). But it transpires that he cannot explain how
this is done without connecting 'S' to communicable rules of a public lan­
guage.
The private linguist maintains that one can give meaning to 'S' indepen­
dendy of any public language, by means of a private ostensive definition. I
have a sensation, and baptize it by concentrating my attention on it and
saying 'S' to myself. Subsequently, I keep a diary in which I record an 'S'
whenever I have the same sensation again. Wittgenstein denies that this
amounts to a meaningful employment of'S'.
'I impress [the sensation] on myself can only mean: this process brings it
about that I remember the connection right in the future. But in the pre­
sent case I have no criterion of correctness. One would like to say: what­
ever is going to seem right to me is right. And that only means that here
we can't talk about 'right'. (PI §258)
This remark has been interpreted as resting on scepticism about memory: I
cannot be certain that I use 'S' only when I have S, because my memory is
fallible. Understood thus, the passage invites the reply that the fallibility of
memory is just as much of a problem in the case of a public language, so
311
 PRIVATE LANGUAGE ARGUMENT
that the argument is either unsound or threatens the possibility of language
in general. Wittgenstein's defenders have retorted that such fallibility is
harmless if mistakes can be corrected, but that this is excluded in the private
case.
Both criticism and defence are right to focus on checkability, but "wrong
to link the latter to scepticism about memory. What is at issue is not the
truth of my utterance 'There's S again', but its meaningfulness. 'There is no
question of my memory's playing me a trick - because (in such cases) there
can be no criterion for its playing me a trick' (LSD 8, see 38-9, 114; PI
§260; MS 166 21), since the original ceremony failed to establish a rule for
the use of 'S'. Stated in general terms, there is no such thing as a private
rule, because a standard of correctness must be checkable (LPP 247). But in
the case of a private standard this is ex hypothesi not the case. It has been
objected that this relies on an indefensible verificationism, since it confuses
the question of whether the private linguist employs a standard of correct­
ness with the question of whether we can know that he does. However,
Wittgenstein argues not that we could not possibly know whether the private
linguist is applying the rule correctly, but that even for him no rule for the
use of 'S' has been laid down. For there is no such thing as a non-opera­
tional standard of correctness, one which cannot even in principle be used
to distinguish between correct and incorrect applications. One might grant
this but insist that while the private linguist's application of 'S' at t\ is incor­
rigible at t\, it can be corrected by him at t . However, justification consists
2 itself, which is as if he tried to measure a ruler against itself or 'to buy sev­
in 'appealing to something independent' (PI §265). Since ex hypothesi, this is
excluded, at ^ nothing distinguishes the private linguist's rectifying a mistake
by reference to a prior rule from his adopting a new rule. Hence, there was
in fact no rule to begin with, but only 'impressions of rules' and a 'pseudo-
practice' (PI §259; MS 180a 76).
Wittgenstein elaborates this general line of argument by arguing that
there cannot be a private ostensive definition, since there are no mental
analogues to the essential features of public OSTENSIVE DEFINITIONS. The logi­
cal category of the definiendum needs to have been determined, that is, 'S'
must be the name of a sensation. However, 'sensation' is a word in our
public language which is defined by reference to behavioural CRITERIA. Since
the private linguist denies or severs this connection, he must explain the
category or 'post' of 'S' afresh. However, simply muttering 'This is S' does
not make 'S' the name of a sensation, since it leaves undetermined what
'this' is. Concentrating one's attention cannot establish criteria of IDENTITY
for subsequent uses of 'S'. Such criteria can be provided only by specifying
what kind of thing is at issue through a sortal term. But the private linguist
has not established what it is he is concentrating on. He cannot say that it is
a certain 'experience' or 'phenomenon', since he lacks the resources for
explaining those terms provided by our public language. He cannot even
312
PRIVATE LANGUAGE ARGUMENT
say that 'S' refers to something he has, since 'has' and 'something' are like­
wise terms from our public language with a determinate grammar. Thus, an
elenctic argument forces the private linguist to the point where he 'would
like just to emit an inarticulate sound' (PI §§257, 261-3; LSD 42, 105; LPE
290). In this vein, Schlick professed his inability even to talk about the 'pri­
vate content' which each person allegedly associates with words (Papers II
306-7). But this is to admit that Wittgenstein is right: one cannot invoke a
private content in philosophical debate.
Even if one granted the putative sample of the private ostensive defini­
tion, there are no ways of checking subsequent employments of 'S' by refer­
ence to it, since nothing determines identity or difference between sample
and described item. There is no established method for comparing sensa­
tions in the way in which there is a method for comparing the lengths of
objects by reference to measurements by a ruler. Moreover, one cannot pre­
serve a sensation for future use as a sample (LSD 42, 110). Investigations §265
considers the suggestion that for such a reproducible sample one might sub­
stitute a mnemonic image of the original sensation. But this procedure is not
like calling up a mnemonic image of a colour-chart. In that case, there is an
independent standard for remembering correcdy. All the private linguist can
appeal to in checking whether he can remember what sensation he asso­
ciated with 'S' is his mnemonic image, which is simply his remembering
what sensation he associated with 'S'. He is checking his memory against
eral copies of the morning paper to assure himself that what it said was
true'. Even if one grants the private linguist the use of the term 'sensation',
all he could mean by 'S' is 'the sensation I'm now experiencing' - not
'such-and-such a sensation I've experienced before'. Consequently, his
'There's S again' does not apply 'S' according to a norm of correctness, and
cannot be a description of something private (PI §§222, 232, 265).
If a private ostensive definition cannot provide a standard of correctness,
its putative sample, the inner object, drops out of the picture as an 'idle
wheel'. In a language-game in which everyone has a box and refers to its
contents as a 'beede', but in which no one has access to the contents of
others' boxes, the contents of the box and their nature are irrelevant to the
meaning of 'beede'. The same holds if we imagine the inner object (the sen­
sation S, the inverted colour-spectrum) changing constantly without our
noticing. The reason is not that the private object is unknowable, but that it
is semantically irrelevant (PI §§271-3, 293, II 207; BB 72~3; LPP 281).
It is tempting to suspect that this leaves us with a kind of BEHAVIOURISM
which denies that there is anything behind the outer behaviour. Wittgen­
stein denies that charge. ' "And yet you again and again reach the conclu­
sion that the sensation itself is a nothing." - Not at all. It is not a something,
but not a nothing either!' The sensation is a semantically irrelevant 'gramma-
313
 PRIVATE LANGUAGE ARGUMENT
deal fiction' only if we construe the grammar of 'pain' on the model of
object and name (PI §§304—8). If we treat 'pain' as the name of a private
object, the question of the identification or misidentiheation of its referent
must arise, since the sensation is supposed to be an object, but cannot be
solved, because it is supposed to be private. There are no criteria of identity
for private mental entities. That does not imply that there are no such
objects, only that even the private linguist himself cannot keep track of
them. Nor does it imply that there are no sensations, but rather that sensa­
tions cannot be understood as private entities. (Equally, the Kant-Strawson
argument that there are no criteria of identity for Cartesian soul-substances
implies neither that for all I know my thoughts might be those of a thou­
sand souls, nor that I do not exist, but only that I cannot conceive of myself
as a soul-substance.)
Words like 'pain', 'itch' or 'tickle' are names of sensations, but not in the
way in which 'table', 'chair' and 'sofa' are names of pieces of furniture. One
can point at a table and say that 'table' is the name of thisns* piece of furni­
ture, but one cannot point at a sensation and say that 'pain' is the name of
thisis* sensation. Instead, to say that 'pain' is the name of a sensation is to
say that there are characteristic behavioural manifestations of pain which
provide criteria for statements like 'She is in pain', and that a sentence like
'I am in pain' is (typically) an AVOWAL - not the report of an inner object
but an expression of pain. In the case of genuine sensations, the problem
about criteria of identity does not arise, since there is no such thing as
either identifying or misidentifying one's own sensations (this is arguably the
point of Investigations §270, which envisages a use for 'S' by treating 'There's
S again' not as a description of private goings-on, but as an avowal).
Even if one does not accept Wittgenstein's alternative account of sensa­
tion-terms, the private language argument undermines the idea that private
experiences provide the foundations of language and of knowledge. It also
implies that to know the meaning of psychological terms we do not need to
have the corresponding sensation or experience. To state meaningfully that
another person is in pain, we do not need a pain, but the concept of pain.
Having the experience does not guarantee mastery of the use of the term.
Equally, someone who applies and explains the term 'toothache' correctly,
but has never had a toothache, knows what 'toothache' means. One might
object that we have no reason to believe that such a person has mastered its
first-person use. But we have if he can say of himself 'I haven't got a tooth­
ache.' The only type of case in which he could be said not to have mas­
tered the first-person use is if he suddenly screams out in agony holding his
cheek, but insists that he hasn't got a toothache (Z §§332-3, 547-8; LSD 9 -
16). This in turn implies the untenability of abstractionism: we could not,
and need not, derive concepts by concentrating on certain features of
experiences and disregarding others. It also reinforces the Kantian attack on
314
PROPOSITION
the myth of the given (not always adhered to by Kant himself, but since
urged by Sellars): the 'preconceptual intuitions' or 'nonconceptual contents'
brandished by empiricists from Locke through the Vienna Circle to Quine
and contemporary theorists of content are at best part of the causal
mechanism underlying speech, and do not feature in the rules which give
meaning to our words. They are semantically and epistemically irrelevant in
that they provide neither the sense of, nor the evidence for, our statements.
proposition Satz means both 'sentence' and 'proposition', a term which
has been variously used to signify what sentences express, the bearers of
truth and falsity, and the objects of propositional attitudes. Wittgenstein's
immediate predecessors all repudiated the idea, shared by idealism and psy-
chologistic logic, that judgements are operations performed on ideas. Frege
distinguished sharply between the sentence, the ideas (Vorstellungen) accom­
panying it, and what it expresses. Every sentence expresses its 'sense', a
'thought', which is neither physical nor mental but part of a Platonic 'third
realm', and is the name of its meaning, which is its truth-value, the True or
the False. Moore treated a proposition as a complex of concepts, which sub­
sists eternally, and is true or fake, regardless of whether it is thought by
anyone. Russell replaced 'concept' by 'term' and held that propositions are
timeless complexes of terms. All three identified true propositions with facts.
But Russell soon came to treat propositions as the complex symbols which
correspond to facts, and the latter as complexes consisting of 'individuals',
the ultimate constituents of reality ('Thought'; Writings ch. 1; Principles ch. 4;
Logic 178-89).
Wittgenstein maintained, against Frege and Russell, that the assertion-sign
is of merely psychological importance (see BELIEF). Logic is concerned only
with the unasserted proposition, which may be a constituent of an assertion, a
question or a command. All propositions can be analysed into elementary
propositions which depict reality by depicting possible states of affairs; the
GENERAL PROPOSITIONAL FORM is: 'Things are thus-and-so' (TLP 4.5). The
Tractatus distinguishes between the propositional sign (Satzzeichen) and the
proposition (Satz): the former is a 'SIGN', a perceptible token-inscription or
-utterance; the latter a 'symbol', a type which is common to all propositional
signs which have the same sense (TLP 3.31-3.32). Tractatus 3.1 states that a
sequence of signs is a meaningful proposition which truly or falsely depicts
reality only if it expresses a thought (Gedanke). Tractatus 4, on the other hand,
states that a proposition with a sense is a thought. The inconsistency is
merely terminological, however. A sentence expresses a thought by virtue,
not of being correlated with an abstract or mental entity, but of having a
projective relation to reality. A thought is simply a sentence-in-use, a propo­
sitional sign in its projective relation to reality (TLP 3. Iff., 3.32ff., 3.5, 4).
The relationship between a propositional sign and a proposition is anal-
315
 PROPOSITION
ogous to that between a dollar bill and a dollar. The bill does not name a
dollar, but to present the bill is to present a dollar. However, another ten­
sion remains, since the Tractatus is committed to a closet mentalism. The
METHOD OF PROJECTION requires that the sense be thought into the proposi­
tion, that is, that the use of the propositional sign be accompanied by a pro­
cess of thinking, more specifically by a THOUGHT, a psychic fact, which is not
identical but isomorphic with the sentence uttered.
'Frege said "propositions are names"; Russell said "propositions corre­
spond to complexes". Both are false, and especially false is the statement
"propositions are names of complexes"' (NL 97). Propositions do not refer
to either a truth-value (Frege) or a complex object (Russell). They have a
different relation to reality. NAMES have MEANING, that is, stand for objects,
propositions have sense, that is, depict a possible state of affairs: 'Names are
points, propositions arrows' (NL 101; TLP 3.143-3.144). In order to under­
stand a name one must-know what it stands for, in order to understand a
proposition one need not know whether it is true (or false), but only what
would be the case if it were true. Wittgenstein connected this insight with
the claim that propositions must be not just bivalent but also BIPOLAR - cap­
able of being true and capable of being false which excludes the possibility
of a proposition's being necessarily true.
Traditional logic maintained that a proposition like 'Plato is the teacher
of Aristode' consists of a subject 'Plato' and a predicate 'is the teacher of Aris-
tode', school-grammar further distinguishes the copula 'is'. Frege and Russell,
by contrast, analyse it into two argument-expressions ('Plato', 'Aristode') and
a two-place concept-word or propositional function 'x is the teacher of y'.
Both Frege's 'concepts' and Russell's 'propositional functions' are patterns of
correlation. The former map arguments onto truth-values, the latter onto
propositions. Having abandoned the view that the subject is united with the
predicate by a mental act (being subsumed under it), Frege and Russell were
at pains to explain what holds together the components of propositions and
facts. Frege accounted for this unity of propositions through a chemical ana­
logy: concept-words (concepts) are 'unsaturated' - they contain a variable -
and hence combine with 'saturated' argument-expressions (objects) to form a
saturated proposition. Russell maintained that among the components of
facts are LOGICAL FORMS, which hold together the components of the com­
plex. But he faced a problem in that aRb and bRa have the same logical
form. His solution was that a and b are linked to R through further rela­
tions, which differ in these two cases, a proposal which invites Bradley's
regress argument against the reality of relations ('Function' 15-17; 'Concept'
197-205; 'Theory' 80-8).
The Tractatus substantially modifies this picture:
(a) Propositions are composed of function and argument (TLP 3.141, 5.47).
316
PROPOSITION
Wittgenstein sided with Russell, in holding that the values of a propositional
function 'fx' are propositions ^fa\ 'fb', etc.), not truth-values. In contrast to
Russell, its arguments are held to be names, not the objects they stand for
(TLP 4.24).
(b) A proposition is a function of the expressions contained in it. In Frege,
the sense (content) of a sentence is a function of the senses of its con­
stituents; in Russell, the proposition itself is the value of the propositional
function. For Wittgenstein, the sense of a proposition is a function of the
meanings of its constituent names (TLP 3.318, 4.024-4.026).
(c) A proposition is 'logically articulated' or composite. It contains two or
more constituents, but is not a mere list of names, since what represents is
not just the assembly of constituents. 'jVbfr "The complex sign 'aRb' says that
a stands in the relation R to b"; but that "a" stands in a certain relation to
"A" says that aRb' (NL 105-6, 96-7; TLP 3.14ff, 4.024-4.032; NB 3.10.14,
28.5./22.6.15). A proposition is a FACT which constitutes a description of a
possible state of affairs.
(d) That propositions are facts also provides Wittgenstein's explanation of
'how the propositional union comes about'. In a sense, all components
(names) are unsaturated, have meaning only in the context of a proposition.
But names combine immediately, without the help of logical glue, just as the
components of states of affairs fit into one another like links in a chain,
without the need of mediating entities or relations (TLP 2.03f., 4.22f.; LWL
120).
According to the PICTURE THEORY, propositions are 'logical pictures' of
reality (TLP 2.18-2.19, 3, 4.03). Their elements - names - stand for ele­
ments of the depicted situation. But what represents is not the propositional
sign itself, but the fact that these names are arranged in a way which, given
a method of projection and the rules of LOGICAL SYNTAX, represents an
arrangement of objects. After the Tractatus the claim that propositions are
facts disappears, and so it should. Like the identification of facts with true
propositions it is a category mistake: unlike facts, propositions can be true or
false, implausible, etc., and are intensional (the proposition that Berkeley was
buried in Christchurch chapel differs from the proposition that the author of
the Principles was buried in Oxford cathedral). What enables a propositional
sign to represent is not the fact that its constituents are combined in a
certain way, but rather that, according to the rules of our GRAMMAR, this
combination can be used to say that such-and-such is the case.
Wittgenstein also came to question the function-theoretic analysis, (i) The
argument/function distinction is merely a sublimation of the subject/
317
 PROPOSITION
PROPOSITION

predicate distinction, (ii) Like the latter, it does not indicate a hidden depth- lack subject and predicate and are not expressed through a propositional
structure discovered by LOGICAL ANALYSIS, but merely a FORM OF REPRE­ clause. This is at odds with what we ordinarily call a 'proposition' or 'sen­
SENTATION, a uniform theoretical mould which we impose on the modey of tence'. But it is noteworthy that our punctuation treats such utterances as
language, thereby concealing the fact that 'there are countless different logi­ sentences. Wittgenstein's account has the merit of illuminating the most
cal forms' (PR 119). (iii) Frege's distinction between objects and concepts is important roles of propositions in our linguistic activities.
too coarse. It papers over the categorial differences between particulars,
numbers, truth-values, circles, places, instants, etc. (PR 120—1, 137). (iv) It is
mistaken to insist that propositions must be complex. One can imagine lan­
guage-games in which expressions cannot be classified into words and sen­
tences - it is mistaken to suppose, for example, that 'Slab!' of Philosophical
Investigations §2 must mean the same as the elliptical sentence 'Slab!' in our
language. Moreover, a non-complex symbol (i.e., name) may be used as a
description (PI §§19-20, 49; BB 77-8). (v) The sense of a token-sentence is
not determined simply by the meanings of its constituents plus its logical
form, but also by how it is used on a particular occasion (see CONTEXTUALISM).
Wittgenstein continued to hold that something like bipolarity defines the
notion of a proposition. Thus, he castigated intuitionism for ignoring that
the law of excluded middle (a precondition of bipolarity) is partly con­
stitutive of what we call a 'proposition' (see MATHEMATICS). However, in other
passages Wittgenstein realized that there is no warrant for restricting the
notion of a proposition to bipolar descriptions of possible states of affairs. It
is legitimate to speak of necessary propositions in mathematics and logic, as
long as one keeps in mind the differences between them and empirical pro­
positions. Not even all empirical propositions fit the narrow picture: the
Welibild propositions of On Certainty (see CERTAINTY) could not simply turn out
to be false; and the law of excluded middle does not hold for certain
counterfactual conditionals, or statements like 'He has stopped beating his
wife' (Z §§677-83; RPP I §§269-74).
In his VERIFICATIONIST phase, Wittgenstein held that only statements about
our immediate experiences qualify as 'genuine' propositions. Later he aban­
doned the idea that propositions display a single general form, and treated
'proposition' as a FAMILY-RESEMBLANCE concept (PI §§23, 65). But from his dis­
cussion of LANGUAGE-GAMES, there also emerges a new uniform conception of
a proposition, albeit a less rigid one. A sentence is a minimal unit for
making a move in the language-game. This conception may have been
partly inspired by Biihler, but starts out from the earlier view that only pro­
positions, not individual words, say or communicate something (a view
shared by Plato, Aristode, Bentham and Frege). It adds the idea that sen­
tences are defined by their role in linguistic activities. Whether or not a
given linguistic form constitutes a proposition does not depend on a certain
structure, but on whether, on a particular occasion, it has been used to per­
form an intelligible linguistic act (PI §50; PLP 317-20, ch. XIII). By this
token, 'Shame!' or 'Ouch!' are propositions, in spite of the fact that they

318 319

R expression of 'a passionate commitment to a system of reference' or of a
realism see ELEMENTARY PROPOSITION; TRUTH
relativism see AUTONOMY OF LANGUAGE; FORM OF LIFE
religion Wittgenstein was a pious man in search of a religion. The young
Wittgenstein seems to have held religion in contempt. A profound change
was brought about by his experiences during World War I. Wittgenstein
tried to prevent 'losing himselT by resigning himself to God's will, and
Christianity is declared to be 'the only safe way to happiness' (GT 16./
25.8./12.11./8.12.14). But this is not conventional theism. Wittgenstein
recognizes that Nietzsche would resist such resignation even at the price of
misery. Moreover, his God is not a personal one, but identical with the
meaning of life, which is also the meaning of the world, and with fate, the
'world which is independent of our will', which is identical with 'how things
stand'. There are 'two godheads: the world and my independent I' (NB
11.6./8.7./1.8.16). One way of reconciling these diverse specifications is this.
Grasping the meaning of life consists in the one godhead - the metaphysical
self - accepting the other godhead - how things are - with Spinozistic
equanimity, because it realizes that how things are has no bearing on the
meaning of life. Hence the resolution to the 'problem of life' is its dis­
appearance (TLP 6.52f). 'God does not reveal himself in the world' (TLP
6.432), that is, in how the world is, but in that it is. Religion is part of the
MYSTICAL realm of value, and hence ineffable. Unlike contingent facts, reli­
gion cannot be expressed in meaningful propositions, it can only be shown.
Unlike the logic of language, but like ethics, it is shown not by meaningful
propositions, but by one's actions and attitudes (see SAYING/SHOWING).
Religion does not occupy an important role in Wittgenstein's later work.
But in his occasional remarks, as well as in lectures and conversations, he
intimates a picture which has been highly influential, and has given rise to
what has been labelled Wittgensteinian fideism (CV passim; LC 53-72; LE
9-10; RW 76-171). He abandoned the idea that religious propositions are
ineffable, and instead made suggestions about their GRAMMAR, the role terms
320
RELIGION
like 'God', 'sin' or 'Last Judgement' occupy in a form of life (M 103-4; LC
71-2; CV 50, 82; PI §373; AWL 32). His reflections deserve the label fide­
ism, since they maintain that religious faith is neither rational nor irrational,
but rather prerational (LC 58-9); and is not assent to a doctrine, but the
certain attitude towards life (CV 64, 85).
On the other hand, by contrast to fideists like Pascal or Kierkegaard,
Wittgenstein is not a Christian apologist. 'I am not a religious man but I
cannot help seeing 'every problem from a religious point of view' (RW 79,
see 93; CV 32-3, 45, 56; LC 63, 70). Wittgenstein does not set out to
endorse a particular religious frame of reference, and repeatedly states that
he does not share any such framework. What he offers instead is a theology
for atheists, an understanding of religion from the outside (as an anthro­
pological phenomenon) which does not accuse it of being either mistaken,
unfounded or nonsensical. This theology involves the following points.
The non-descriptive and non-cognitive nature of religion Religious statements do not
describe any kind of reality, empirical or transcendent, and do not make
any knowledge claims (LC 59-63). Someone yho believes in a Last Judge­
ment does not use expressions like 'Such and such will happen' to make a
prediction, but rather to express a commitment to a 'form of life', for exam­
ple one in which people feel constantly admonished by God's approval or
disapproval. Indeed, if he were making a prediction, it would not count as a
religious belief (LC 56-8; CV 87). Someone who believes in an afterlife is
not committed to the Cartesian notion of a soul-substance, but only to a
certain picture, although Wittgenstein sometimes admits that he does not
have 'any clear idea' of what the picture amounts to (LC 70-1; PI II 178;
RPP I §586). The belief in miracles is a propensity to be impressed by
certain coincidences. Someone who says 'It is God's will' may be uttering
something similar to a command like 'Don't complain!' (CV 61).
The existence of God By the same token, religious terms like 'God' do not
refer to entities, and to state that God exists is not to make a statement
about the existence of a certain entity (LC 63; CV 50, 82). It expresses a
commitment to a certain frame of reference or a form of life, a commitment
which is brought about not by argument but by a certain upbringing or
certain experiences.
JVo need for proof Wittgenstein regarded attempts to prove the existence of
God or the immortality of the soul as based on misunderstandings of reli­
gious belief, and of the roles of statements concerning God or the afterlife in
the form of life of religious believers (RW 107-8; CV 82~5; GB 119). This
holds not just for proof by a priori argument, but also for proof by revela­
tion or divine inspiration. To say that something has been 'revealed' to me
321
 RELIGION
by God is not to specify a source of knowledge, but to state a decision (OC
§§361-2). Equally, Christian belief does not rest on the historical truth of the
Gospels; our attitude towards these stories is different to our attitude
towards a historical account (CV 32).
Religion as a sui generis form of discourse Although he characterizes praying as
a language-game (PI §23), Wittgenstein nowhere states that religion con­
stitutes a separate form of life. But the idea that it constitutes a sui generis
grammatical system is implicit in the idea of the non-descriptive and non-
cognitive nature of religion. Like any such system, religious discourse is
AUTONOMOUS, it neither corresponds nor fails to correspond to an 'essence of
reality'. Scientific discourse shows neither that religious discourse is epistemi-
cally unfounded, nor that it is meaningless, it merely expresses a different
attitude towards the world and life (CV 5; MS 134 143-6). Like Kant, Witt­
genstein tries to preserve religion from the encroachment of science and
metaphysics. Unlike Kant, he suggests that it is also independent of ETHICS.
Faith vs. superstition Wittgenstein distinguished between religious faith on the
one hand, and superstition on the other. Throughout his career, Wittgen­
stein characterized superstition as the false belief in supernatural causal
mechanisms, 'a sort of false science' (TLP 5.1361; OL 31; CV 72; GB 111-
31). However, he also uses the term 'superstition' for any attempt to justify
religious beliefs through some kind of evidence. He condemns both 'Russell
and the parsons' for having tried to assess the rational credentials of reli­
gion, and contrasts the grey wisdom of philosophy with the colour of life
and religion (LC 57-9; RW 102; CV 22, 62, 86). Just as metaphysics is mis­
guided philosophy, superstition is misguided religion, in both cases because
non-factual propositions are understood as statements of super-empirical
facts. Genuine religious beliefs and rituals are expressive rather than instru­
mental (see ANTHROPOLOGY). What gives them their meaning is not empirical
or metaphysical beliefs, but their role in the practice of the believer (CV 85).
While these points contain important insights into the workings of religious
language, it is doubtful whether Wittgenstein's conclusions are justified. It is
difficult to avoid the impression that double-standards are being employed
when one is told, for example, that while '1 = 1' is a 'perfectiy useless' pro­
position (see IDENTITY), '1 person + 1 person + 1 person = 1 personal God'
makes perfectiy good sense in the mouth of a trinitarian Christian. Wittgen­
stein's reply would be that by contrast to metaphysical pronouncements,
religious propositions have a genuine use. However, even the pronounce­
ments of revisionary metaphysics are used in a form of discourse. Wittgen­
stein would reply that unlike religious propositions, they are not embedded
in a form of life, that is, make no difference to our extralinguistic activities.
But some philosophical views (e.g., those of Marxism) have had a profound
322
RULE-FOLLOWING
influence on the practice of their adherents. Furthermore, it is unclear how
a doctrine like that of the Trinity can be given a sense simply by being
uttered as part of religious practices. Finally, it is problematic to say that it
is the religious practice, for example in a rite, which gives content to reli­
gious doctrines, because the doctrines themselves are supposed to underpin
the practice. Although a belief in the existence of God may differ from a
belief in, say, the existence of quarks, it cannot simply amount to commit­
ting oneself to a religious life, since the belief will typically be part of the
reason for making such a commitment.
There are two replies to this attack. First, the fact that believers justify
their beliefs by reference to evidence or religious experience no more shows
that the former rest on the latter than the fact that people speak of thinking
as a brain state confirms materialism. What counts is not believers' philo­
sophical prose, but their practice, and 'the religious practices of a people'
are not abandoned when their evidence is shown to be inadequate (LC 60-
2; GB 121). But while a people as a whole rarely abandon their religious
beliefs, reflective individuals certainly do, if they come to the conclusion that
their justifications have been inadequate. Wittgenstein is committed to the
view that such individuals are confused, while th*se who persist in citing
what he (without argument) regards as flimsy justifications are not. He is
also committed to the view that rationalist philosophers of religion (such as
Aquinas, Kant, and Kenny) are superstitious. What implies these unpala­
table conclusions is the following line of reasoning: if religious beliefs were
based on evidence or metaphysical doctrines, they would be 'stupid'; reli­
gious beliefs are not stupid; hence they cannot rest on evidence or metaphy­
sical doctrines (GB 119, 125). But religious beliefs, like philosophical
doctrines, may not simply be stupid (may reflect deep-rooted human aspira­
tions and be based on profound reflections) and yet involve irrational or
incoherent beliefs.
The second reply is that whereas causal beliefs explain and justify certain
instrumental practices, religious doctrines do not explain or justify the ritual
practices, since the explanations make sense only within the context of those
practices (see GB 121-3, 129) - for example, the doctrine of the Apostolic
succession belongs to, rather than explains, the Catholic Mass. But while the
doctrine may not provide a reason for participating in the Mass for some­
one who does not share the religious way of life, it is nevertheless the believ­
er's justification, and it explains the Mass: we can understand why a belief
in the Apostolic succession provides a reason for participating in the Mass,
even if we find the doctrine incoherent and the practice irrational.
rule-following Rules play a crucial role in Wittgenstein's philosophy
because of two abiding convictions: firstly, language is a rule-guided activity;
secondly, the a priori status of logic, mathematics and philosophy derives
323
 RULE-FOLLOWING
from such rules. In the Tractatus, linguistic rules form LOGICAL SYNTAX, a com­
plex calculus of inexorable norms hidden behind the surface of natural lan­
guages. By the mid-thirties, Wittgenstein had moved away from this
CALCULUS MODEL of language. In particular, he had rejected the idea of rules
which guide linguistic behaviour and determine what it makes sense to say
without being known to us. To clarify the way in which rules guide our
behaviour and determine the meaning of words is the strategic role of his
celebrated discussion of rule-following. Because of the connection with
linguistic meaning, understanding and logical necessity, the topic is central
to his philosophy of language, philosophical psychology and philosophy of
mathematics (it pervades Remarks on the Foundations of Mathematics).
Wittgenstein did not try to provide an analytic definition of 'rule', since
he considered it a FAMILY-RESEMBLANCE concept best explained by reference to
examples. But several points can be culled from his remarks (WVC 153-4-
AWL 153-5; BT 241', PG 117-18; BB 90-8; RFM 321; SDE 24; PLP 8 2 '
137-44).
(a) Rules are standards of correctness; they do not describe, for example,
how people speak, but rather define what it is to speak correcdy or mean­
ingfully.
(b) There is a difference between a rule and its expression, a rule-formula­
tion, just as between a number and a numeral (e.g., the same rule can be
expressed in different languages). But the difference is not one between an
abstract entity and its concrete name, but one between a normative func­
tion, and the linguistic form used to perform that function. We can clarify
the notion of a rule by investigating the role of rule-formulations.
(c) Unlike commands or orders, rules are inherendy general in that they
govern an often unlimited multiplicity of occasions.
(d) Features like (a) or (c) are not tied to particular forms of words - a
.GRAMMATICAL proposition expressing a linguistic rule need not be a meta­
linguistic statement about the employment of words, or contain expressions
of generality. Rather, they depend on whether an expression has a norma­
tive function on a given occasion.
(e) 'Rule-following' is an achievement-verb: there is a difference between
believing that one is following the rule and actually following it.
(f) The crucial point for the change in Wittgenstein's conception of linguistic
rules is that there is a difference between following a rule and merely acting
in accordance with a rule. Although rule-following presupposes a regularity
in behaviour, this does not distinguish it from natural regularities like the
movement of planets or human acts which happen to conform to a rule
324
RULE-FOLLOWING
unintentionally. If an agent follows a rule in Oing, the rule must be part of
his reason for <E>ing, and not just a CAUSE. He must intend to follow the rule.
However, this intentionality is only virtual. He does not have to think about
or consult the rule-formulation while <&ng, it is only required that he would
adduce it to justify or explain his Oing. This excludes the idea of rules
which are completely unknown to the agent (as those invoked by the calcu­
lus model). It also sets rule-following apart from 'inspiration'. These are
cases in which the agent is guided passively, without being able to explain
why he acts as he does, or to teach others the technique of following this
guidance (PI §§207-8, 222, 232; BB 12-14; RFM 414-22).
The notion of rule-following thus explained poses two interrelated prob­
lems. One concerns our UNDERSTANDING of rules, the other the normativity of
rules. In Philosophical Investigations §§143-84, Wittgenstein attacks the idea that
understanding is a mental state or process from which our application of
words flows. In §§185-242, he turns to the question of how a rule deter­
mines what counts as a correct or an incorrect application. We distinguish
between rules which do determine an answer at each step (e.g., 'y = 2x') and
those which do not (e.g., y # 2*') (PI §189; RFM 35-6). But even in the
former case a puzzle arises. I teach a pupil the arithmetic series '+2'. But
when he first proceeds beyond ljOOO, he says '1,000, 1,004, 1,008'. The
question is, what are our grounds for saying that he has misunderstood the
rule? A rule like '+2' covers an unlimited number of steps. But all the pupil
has to go by is a rule-formulation, and a few steps as exemplification. How
does the rule-formulation, a mere sign, manage to determine in advance an
unlimited number of steps? Wittgenstein considers four different answers to
these puzzles.
Mechanism Understanding a rule is a disposition, and statements about dis­
positions are ultimately statements about a mechanism (AWL 83-4, 91), in
this case a mental or neurophysiological one which produces the right
actions in the appropriate circumstances. This picture turns the rule from a
reason for action into a cause, and, as a result, violates the normative nature
of rules. The relationship between a mechanism and its causal consequences
is merely external, at the mercy of contingent facts. To say that '1,002' is
the next correct step is not to say that it is one which, say, a computer will
take, or the one which I am predisposed to take. Indeed, we judge rather by
reference to the rule whether the computer has broken down, or whether I
am inclined to give the right answer (see PI §§149, 220; RFM 332). Equally,
that I meant the pupil to write '1,002' is not a counterfactual 'Had you
asked me which number he should write,...'; unlike the former, the latter is
an empirical hypothesis.
325
RULE-FOLLOWING
Platonim The rule, unlike its linguistic expression, is an abstract entity
which somehow already contains the whole series of even numbers. This
replaces a puzzle by a mystery. It is unclear how the mind grasps such enti­
ties. To do justice to the normative nature of rules, Platonism invokes a
'super-strong connexion' which is not just causal. The rule is a 'logical
machine', an 'ethereal mechanism' which is immune to break-downs and
churns out an infinite totality of applications independendy of us, or a set of
rails which drag us along inexorably. But this illegitimately 'crosses different
pictures'. Both mechanism and Platonism run together 'being determined by
a matter of fact' with 'being determined by a stipulation'. In claiming that
'1,002' is the next correct step, we do not predict that people or machines
will arrive at this result, but rather stipulate that the rule has not been fol­
lowed if they don't (PI §§191-7, 218; RFM 83; LC 14-15; AWL 83-4; Z
§375; MS129 176; TS219 33).
Mentalism The pupil's going wrong consists in his failing to intuit what I
meant by my instruction. In a way, this is correct (PI §190). But I did not
think in advance of this particular step, and could not have thought in
advance of the infinite number of steps which constitute the series (PI §186).
Even if the rule were a mental state or an abstract entity, that would not
explain rule-following, since the question remains of how that rule is to be
applied, what the METHOD OF PROJECTING it onto reality or translating it into
action is. The idea of a self-applying rule which already 'contains' all the
steps to be taken in advance is simply mythological (PI §§195-7; BB 142;
AWL 89-90, 131-5).
Hermeneutics How I meant the instruction is expressed in how I interpret it.
This suggests that while the rule-formulation by itself does not determine the
next correct step, its interpretation does. Against this, Wittgenstein invokes a
regress argument which goes back to Kant (A133-4). An 'interpretation' is
'the substitution of one expression of the rule for another' (PI §201), and
hence does not get us closer to the correct application. One might think
that this definition of interpretation is too narrow (one can interpret without
actually substituting one expression for another). But this would not impugn
the regress argument. If an interpretation is supposed to provide the sym­
bolic content missing from the mere rule-formulation, it must be 'something
given in signs', 'a new symbol added to the old one', which leaves rule plus
interpretation 'hanging in the air' (see LWL 24; PG 47; BB 4-5, 33-5, 124;
Z §§229-35; PI §§84-7, 198). Rule-following cannot be explained by invok­
ing rules on how to apply rules.
To some commentators, Wittgenstein's rejection of these explanations has
suggested that he adopted a rule-scepticism. There is no such thing as
326
RULE-FOLLOWING
objectively following or violating a rule, since any action is, on some inter­
pretation, in accord with the rule. Even I myself cannot know what I meant
by the instruction 'Add 2.' It would follow that there is no such thing as
linguistic meaning. But since this semantic nihilism is obviously self-refuting,
some difference between correct and incorrect proceeding is reintroduced
through a Humean 'sceptical solution': what makes the pupil's proceeding
incorrect is that it is rejected by the linguistic community. Alas, this is no
solution, sceptical or other. If I cannot know what / mean by 'Add 2', the
pupil has no chance of knowing whether in saying 'Carry on!' or 'Look
what you've done!' (PI §185) the community means to accept or reject his
steps.
Wittgenstein struggled hard to steer a fine between the Scylla of rule-
scepticism, and the Charybdis of pseudo-explanations like the four noted
above. There are occasional hints of rule-scepticism in his reflections on
FAMILY RESEMBLANCE and on MATHEMATICAL PROOF. Furthermore, some passages
agonize over the 'gap' between a rule and its application, and contemplate
whether a new decision is required at each step (PI §§186, 198; MS 180a
68-75; MS 129 117, 182). But he also stated that, the sceptical 'paradox'
according to which 'no course of action could be determined by a rule',
because there is always another possible interpretation, is based on a 'mis­
understanding'; it shows not that the rule leaves its applications un­
determined, but that 'there is an understanding of the rule which is not an
interpretation, but which is exhibited in (what we call "obeying the rule" and
"going against it" in actual cases' (PI §201). Some passages suggest that the
gap between a rule and its application is crossed in our practice (TS211
112). This is correct in so far as rule-following is essentially a practice (PI
§202; RFM 335, 344-6). But to think that 'a gap exists between a rule and
its application' is a 'mental cramp' in the first place (AWL 90; LSD 24).
The relation between a rule and its correct application, like other INTEN­
TIONAL relations, is INTERNAL. TO understand the rule is to know how to
apply it, to know what counts as acting in accord with it and what violating
it (RFM 331-2).
Any finite sequence of numbers (e.g., 1, 4, 9, 16, 25) is compatible with
an infinite number of mathematical series. By the same token, any finite
array of behaviour is compatible with 'any number' of rules (BB 13). It fol­
lows that extrapolating a rule from its extension, that is, behaviour described
without reference to the rule, is underdetermined (this converges with
Quine's thesis of the indeterminacy of translation). But it does not follow
that the rule leaves its application underdetermined, as rule-scepticism has it.
Internal relations are de dicto, they depend on how we describe things: if the
rule-formulation 'Add 2' and the utterance '1,000, 1,002, 1,004' are descri­
bed phonetically, it is no more possible to tell whether the latter is a correct
application of the former than to derive the age of a ship's captain from its
327
RULE-FOLLOWING
dimensions (TS211 494). But the rule provides a standard for describing
actions as 'obeying' or 'going against it'. Between the rule and its applica­
tion there is a category difference (as between an ability and its exercise),
but no gap to be bridged: if '1,002' were not the next correct step, this
series would not be what we call 'the series of even numbers'.
This seems to reintroduce the sceptical solution. But by contrast to the
latter, what counts as a correct application of a word or rule is precisely not
determined 'by the agreement of human beings' but rather by reference to
the rule itself. Communal agreement in judgements is a FRAMEWORK condi­
tion: unless we could agree on how to apply a rule, we would not engage in
that practice (RFM 406; Z §§319, 428-31; PI §§219, 241; RPP II §414). But
'applying the rule correcdy' does not mean 'doing what most people do'
(rules can be misapplied by a whole group, as with the first Newtonian cal­
culations of the moon's orbit). The rule is our standard of correctness. But
nothing is such a standard unless it is used as such. There is no rule unless
there is a practice of calling such-and-such 'obeying' or 'going against it'.
Internal relations are effected by our normative activities - we teach and
explain rules, and criticize, justify or characterize actions by reference to
them (PI §§197-202; PG 213-14; RFM 344-5; LFM 83).
Unlike the connection between the proposition that p and the fact that p,
that between an arithmetic formula and its application is not direcdy visible.
So how does the pupil know what we mean? Through our explanations and
instructions! If 'Add 2' meant 'Add 2 up to 1,000, 4 from 1,000 up to
2,000, etc.', it would not be correcdy explained by '0, 2, 4, 6, and so on'.
Given our arithmetic techniques, there is an onus to specify that the pattern
changes after 1,000. But 'who says what "change" and "remaining the
same" mean here?' (RFM 79-81). Appearances notwithstanding, such pas­
sages do not express a rule-scepticism, but make two other points. Firsdy,
although the rule does not leave us in the lurch, we are not logically com­
pelled to follow one rule rather than another. The rule does not grab the
pupil by the throat. If he continues '1,004', he is simply not playing our
game. What he does is not what we call 'adding 2', but again there is no
logical compulsion why we must use the signs 'adding 2' in this way (AWL
88-9; RFM 35-8, 328-9, 414; LFM 108, 183-7). Secondly, the notions of
rrjENTTTY ('doing the same') or of 'accord' do not provide us with an inde­
pendent ground for reprimanding a deviant pupil (PI §§214-16, 223-7; BB
140; Z §305; RFM 348-9, 392-3, 405). The concepts of 'ride-following' and
of 'doing the same' are interwoven. It is a grammatical proposition that
doing the same thing as before is correct if what was previously done is cor­
rect. But what counts as 'doing the same' can only be determined by refer­
ence to a particular rule. Our deviant pupil is doing the same as before
relative to the rule 'Add 2 up to 1,000, 4 up to 2,000, 6 up to 3,000, etc'
We can interpret any rule-formulation in different ways, and have to
328
RULE-FOLLOWING
interpret it in some way if it is ambiguous or unclear. Yet this means not
that interpretation goes on for ever, but only that what for us is the last
interpretation need not be. 'Interpretation comes to an end' (RFM 341-2;
PG 147; BB 34; PI §201). I can justify my writing '1,002, 1,004,...' by
reference to my instructions (e.g., by counting '1,001, 1,002 - that's adding
2 to 1,000'). Asked why I understand the latter in this way I can only say
'This is simply what I do', because I have exhausted the justifications and
reached 'bedrock'. But the brute factuality of our practice does not provide
room for scepticism. Acting 'without justification' after all justifications have
been given is not acting 'wrongfully' (PI §§211, 217, 289, 381; RFM 199,
406; Z §§300-2). If a sceptic doubts that the rule-formulation 'Add 2' as we
use it demands '1,002' at the 501st step, he cannot be talking about those
relata, which are defined by this fact. His doubt cannot even address the
grammatical proposition it purports to address.
Wittgenstein describes rule-following as a social practice, speaking of 'cus­
toms', 'habits' and 'institutions' (PI §199). The question is whether he held a
community-view, according to which rule-following is possible only within a
social community. This is suggested by the claim that 'it is not possible to
follow a rule "privately"' (PI §202), but the scare-quotes may indicate that
what is meant is not rules which are unshared but only those which are
unsharable in the sense of the PRIVATE LANGUAGE ARGUMENT (PI §256), in
which this passage originally occurred. Investigations §199 denies that there
could be 'only a single occasion' (nur einmal) on which 'only a single person'
(nur ein Mensch) followed a rule. But it has been replied that this is due to the
fact that the Investigations discusses rules which are communal; while insisting
that what is essential to rule-following is only a multiplicity of occasions, not
a multiplicity of speakers (PI §§204-5; RFM 334-6, 346; Z §568). Both parts
of this reply would be contrived were it not for two facts: firsdy, there is no
plausible rationale for restricting rule-following to a community; secondly,
the Nachlass explicitly condones the possibility of a solitary person like
Robinson Crusoe following and inventing rules. Proponents of the commu­
nity-view respond that this is possible only because Crusoe had been a
member of a community. But Wittgenstein righdy insists that whether some­
one follows rules depends on what he is capable of doing, not on how he
acquired that capability (MS 124 213-21; MS 165 103-4; MS 166 4; PG 188;
BB 12, 97; PI §495). Yet, Wittgenstein also suggests that rule-following is
typically social, and that some rule-guided activities - including not just those
which are communal by definition like buying and selling, but also doing
mathematics — require the context of a social and historical 'way of living'
(RFM 335-50; PI §§200-5, 337). One can play even patience only if the
institution of the game exists.
329

s that there is no soul (TLP 5.5421), the truth in SOLIPSISM - that the 'world is
saying/showing The distinction between what can be said by meaningful
propositions and what can only be shown pervades the Tractatus from the
Preface to the famous final admonition 'Whereof one cannot speak, thereof
one must remain silent.' To Russell, Wittgenstein wrote that it was 'the
main point' of the book, and 'the cardinal problem of philosophy' (RUL
19.8.19). In a letter to von Ficker (FL 10./11.19), he proclaimed that the
Tractatus 'consists of two parts: of the one which is here, and of everything
which I have not written. And precisely this second part is the important
one. For the Ethical is delimited from within... by my book; and I'm con­
vinced that, strictly speaking, it can ONLY be delimited in this way.'
There are reasons for regarding this as a slightiy hysterical piece of self-
promotion vis-a-vis a potential publisher. Although Wittgenstein harps on the
existential insignificance of the sayable (TLP Pref., 6.41-6.522), his work
before and after concerns not what the Tractatus leaves unsaid, but what it
tries to say. The Tractatus has indeed two parts, a logical one (atomistic ontol­
ogy, picture theory, tautologies, mathematics, science), and a mystical one
(solipsism, ethics, aesthetics). The real significance of the saying/showing dis­
tinction lies in the fact that it holds the two together by proscribing both
propositions about the essence of symbolic representation and mystical pro­
nouncements about the realm of value. It is the cardinal problem of philoso­
phy because it transformed Wittgenstein's conception of the subject it can
no longer be the doctrine of the logical form of propositions (NL 106), since
logical form cannot be stated. Instead, it is a clarificatory activity which sig­
nifies 'what cannot be said, by presenting clearly what can be said' (TLP
4.115).
Wittgenstein's index of ineffabilia includes the pictorial form common to
picture and what is depicted (TLP 2.172-2.174), the meaning of signs and
that two signs have the same meaning (TLP 3.33fT., 6.23), that a given
symbol signifies an object or a number (TLP 4.126), the sense of a proposi­
tion (TLP 4.022, see 2.221, 4.461), the logic of facts (TLP 4.0312), the logi­
cal multiplicity or form of a proposition and of reality (TLP 4.041, 4.12f.),
that a proposition is about a certain object (TLP 4.1211, 5.535), that some­
thing falls under a formal concept (TLP 4.126), that logical propositions are
330
SAYING/SHOWING
TAUTOLOGIES and do not refer to LOGICAL CONSTANTS (TLP 4.0621, 4.461),
that one proposition follows from another (TLP 5.12-5.132, 6.1221), the
limits or scaffolding of language and the world (TLP 5.5561, 5.6f., 6.124),
my world' (TLP 5.62), that there are laws of nature (TIP 6.36), the ethical
and eveiything that is 'higher' (TLP 6.42L), the meaning of life - the mys­
tical (TLP 6.52ff.), the pronouncements of the Tractatus itself (TIP 6.54).
One can distinguish the following clusters:
(a) the logical form common to propositions and what they depict (inex-
pressibility of the harmony between thought and reality);
(b) the meaning of signs and the sense of propositions (prohibition of
semantics);
(c) the logical relations between propositions (no rules of LOGICAL INFER­
ENCE);
(d) the logico-syntactical category of signs (formal concepts are pseudo-
concepts);
(e) the structure of thought and world (limits to thought are set from
within);
(f) the mystical (the ineffability of value).
Expressions which try to state any of the above are 'pseudo-propositions'.
What unites them is the contrast with the BIPOLAR propositions of science.
While the latter make factual statements, depict combinations of objects that
may or may not obtain, the former attempt to say things that could not be
otherwise. It might seem that being necessary is not a sufficient condition
for being a pseudo-proposition, since tautologies and contradictions are not
pseudo-propositions. But this is due to the fact that the latter are degenerate
propositions produced by licit combinations of genuine propositions (NM
118). Pseudo-propositions do not depend on how things are, since they con­
cern 'transcendental' preconditions of representation and the world (NB
24.7.16; TLP 6.13, 6.421). It is unclear why (f) should have this transcen­
dental status. Moreover, what can be shown cannot be said (TIP 4.12 Iff.).
And in the case of (a)-(e) the reverse also holds. What such propositions try
to say is shown by bipolar propositions and their limiting cases - tautologies
and contradictions. But there are no meaningful propositions which even
show, for example, ETHICAL value. Unlike the logical, the mystical is trans­
cendent, not just transcendental.
Leaving aside this special case, the underlying idea is that the precondi­
tions of symbolic representation, the rules of LOGICAL SYNTAX, cannot them­
selves be represented (NM 108-9). They cannot be represented by bipolar
propositions, because they concern essential features which language and
reality must share for the former to represent the latter. Yet the saying/
331
 SAYING/SHOWING
showing extinction is not simply based on a dogmatic stipulation that only
bipolar propositions make sense. Rather, the principle of bipolarity is itself
informed by insights into the peculiar nature of attempts to state essential
features of symbolism. For one thing, unlike bipolar propositions, such pro­
positions exclude not a genuine possibility, but rather something which con­
travenes logic, and hence the bounds of sense. But the attempt to refer to
something illogical, even for the purpose of excluding it as NONSENSE, is itself
nonsensical - this is the point of (e). For another, no proposition can say
something about the logical properties of language: either such a proposition
itself conforms to logic, then those logical properties must already be under­
stood (circularity), or it does not, then it cannot be a meaningful proposition
(an illogical language is impossible) (TIP 3.031, 4.12, 5.4731).
This general point is applied by (a) to the picture theory. It is not a
dogmatic exclusion of. self-referential propositions. Nor is it a matter of the
impossibility of a proposition or model's depicting how it depicts. If a fixed
map, for example, depicted itself (on a smaller scale), together with its key,
this would lead to a regress, since the key to reading the key would have to
be provided as well. But this is the impossibility of a picture's depicting its
own METHOD OF PROJECTION. A picture cannot depict its own 'pictorial form',
the possibility of structure which it must share with what it depicts, for a dif­
ferent reason, namely that it cannot represent it as a possibility. For the
pictorial form of a proposition is one of its 'internal properties' (TLP 4.122-
4.1241) - it could not lack that pictorial form without ceasing to be the
picture it is. By the same token, no other proposition could represent it as a
possibility, which means that there can be no bipolar propositions about the
pictorial forms of propositions.
It has been suggested that the saying/showing distinction derives from
Frege's paradox of concepts. Frege distinguished sharply between objects or
arguments, which are saturated, and concepts or functions, which are un­
saturated - i.e., cannot stand on their own but demand completion
through an argument. This led Frege to make the paradoxical statement
that 'the concept horse is not a concept.' For in attributing properties to a
concept we have to use a name ('the concept horse') to refer to something
which is unsaturated, although names can only refer to saturated entities.
Frege's paradox arises out of the untenable idea that concept-words ('is a
horse') name unsaturated entities and that names ('the concept (of a) horse')
cannot perform that role because they do not reflect the unsaturated
nature of what they try to refer to. It involves a profession of linguistic
impotence, since the attempt to refer to concepts through names is a mis­
take forced on us by language ('Concept' 195-9; Posthumous 193). But it is
not the seed of the saying/showing doctrine. Even (d), which deals with
'concepts', prohibits, not referring to unsaturated functions by names, but
any use of formal (i.e., categorial) concepts, which include 'name', 'object',
332
SAYING/SHOWING
'colour' and 'number', as in 'A is an object' or '1 is a number' (TIP
4.126-4.1274).
One might respond that the Tractatus extends Frege's point because it
holds that all NAMES, including those of objects, are unsaturated. But while
Frege is worried about referring to unsaturated entities, the Tractatus is wor­
ried about predicating of a symbol that it belongs to a logico-syntactic cate­
gory. This worry arose out of reflections on Russell's theory of types (RUL
1.13; NL 96-101). Russell prevents the set-theoretic paradoxes by prohibit­
ing sentences which predicate of a thing of one logical type (e.g., classes)
properties which can be predicated only of things of a different type (e.g.,
individuals). It might prohibit, for example
(1) The class of lions is a lion
through a rule like
(1') 'The class of lions is a lion' is nonsensical.
According to Wittgenstein, such a theory is neither possible nor necessary.
(1') is either about SIGNS - in which case it states a contingent fact about
arbitrary conventions, not a logical rule. Or it is about symbols. In that case
it must refer to the sense or meaning of expressions. But it cannot refer to
the sense of (1), which is ex hypothesi nonsensical. Nor can it refer to the
meaning of the names which ultimately constitute (1). For those constituents
do not have a meaning in advance of their logical syntax being fixed.
Therefore the rules of logic cannot be expressed through propositions of the
form ' " 4 " must have such-and-such rules because it refers to an object of
such-and-such a type' (TLP 3.33ff.) - this is the point of (b).
The point of (d) is that we cannot talk about either the logico-syntactical
category of a name or the ontological category of an object with the help of
formal concepts. The ontological category of an object is determined by its
LOGICAL FORM, that is, by what other objects it can combine with in a state of
affairs. That A is a visual object means that it can combine with colours but
not with a pitch (TLP 2.0251; PT 2.0252). But the form of an object can
neither be named (it is not itself an object) nor be described through a
formal concept like 'colour'. Rather, it is shown by the fact that its name is
a substitution instance of a given kind of 'propositional variable' (TLP
4.127ff.). If we replace one of the 'constituents' of
(2) A is red
by a placeholder, we get a propositional variable, or propositional function
(Russell)
(2') X is red.
333
 SAYING/SHOWING
The variable is given by the determination of its values, that is, by stipu­
lating what sort of propositions can be constructed through filling the
argument-place (TLP 3.3Iff.). The values of (2') are all those propositions
we get by substituting a name for X; the variable 'collects' all the proposi­
tions of the form - 'A is red', 'B is red', etc. The formal concept of a visi­
ble object is given by this variable; it is the constant form of all expressions
which can be meariingfully substituted in (2'). In an ideal notation there
would be a distinct variable and a distinct style of names for each logical
category.
A material concept like 'red' can occur in a genuine proposition like (2),
but a formal concept like 'visible object' cannot. For it is in effect a variable,
and a proposition can contain only apparent (i.e., bound) variables (see
TAUTOLOGY). The second step above is trivial - (2') is not a proposition but a
propositional function. But the insistence that a formal concept is in effect a
variable is once more based on the idea that there can be no propositions
ascribing INTERNAL properties to things. 'A is an object' or 'Red is a colour'
are pseudo-propositions, but what they try to say is shown by properly ana­
lysed empirical propositions in which 'A' or 'red' occur. This is the logical
core of the saying/showing distinction: although the rules of logical syntax
cannot be expressed in philosophical propositions, they show themselves in
the logical structures of non-philosophical propositions.
Wittgenstein claims that his theory of symbolism can replace the theory of
types because Russell's paradox concerning the set of all sets which are not
members of themselves is disposed of by realizing that a propositional func­
tion cannot be its own argument (TLP 3.332f.; NL 96, 107). That last claim
follows from Wittgenstein's conception of a propositional function (which in
this respect resembles Frege's conception of concepts). If a function could be
its own argument, there would be a proposition '/(fay. However, in such a
construction the inner f must refer to a function of the form Ox, the outer
to one of the form *P(Ox). The two have the sign f in common, but neces­
sarily different meanings, that is, are different symbols, simply because noth­
ing can be a proper part of itself. It follows that one and the same
propositional function cannot occur twice in a proposition, and hence that
self-predication is impossible. Ruling out self-predication prevents the propo­
sitions which yield Russell's paradox - 'xex' and 'x£ x' - if classes are (as
Russell held) logical fictions such that 'e' is explained through predication:
self-membership is a case of self-predication, and hence ruled out.
An immediate consequence of the saying/showing distinction is that the
propositions of the Tractatus itself are nonsensical, since they employ formal
concepts ('fact', 'proposition', 'object') to make claims about the essence of
representation. Notebooks 20.10.14 suggests that such pseudo-propositions at
least show what they try to say. But unlike tautologies, which show the
structure of the world, philosophical pseudo-propositions cannot show any-
334
SAYING/SHOWING
thing, since they do not employ symbols in a meaningful way. The Tracta­
tus's penultimate remark accepts this conclusion:
My propositions serve as elucidations in the following way: anyone who
understands me eventually recognizes them as nonsensical . . . (He must,
so to speak, throw away the ladder after he has climbed up it.) He must
transcend these propositions, and then he will see the world aright. (TLP
6.54; for the ladder-image see Mauthner, Beitrdge I 2 and Schopenhauer,
World n ch. 7; it is later repudiated, see MS 109 207)
The Tractatus is committed to distinguishing nonsense which is based on mis­
understanding logical syntax from 'important nonsense' (Ramsey, Mathematics
263) which is based on a correct insight into logical syntax, and tries to say
what can only be shown. If, as some have claimed, its pronouncements were
meant to be nonsense in the first sense, it would be neutral between, for
example, Frege's and Russell's idea that propositions are names of objects
and Wittgenstein's own idea that they are facts, which is obviously not the
case. The Tractatus is neither an existentialist joke, nor a protracted nonsense
poem with a numbering system. It was intended as the swansong of meta­
physics and violates the bounds of sense only to attain the correct logical
point of view which allows one to engage in critical logical analysis without
any further violations (TIP 4.1213, 6.53).
The saying/showing distinction is a response to a problem facing any
attempt to identify the bounds of sense with the limits of empirical knowl­
edge, namely that establishing such bounds is not itself empirical (note
Kant's difficulties in avoiding knowledge claims about things as they are in
themselves). It is heroic, but self-defeating. As Ramsey pointed out, it resem­
bles the child's remarks in the dialogue: A: Say 'breakfast'!; C: I can't. A:
What can't you say? C: Can't say 'breakfast'. One may therefore sym­
pathize with Russell's suggestion that this impasse might be overcome by
talking about the logical properties of our language in a meta-language
('Introduction'). In The Logical Syntax of Language, Carnap elaborated this idea.
He suggested that the limits of language can be expressed by switching from
propositions in material mode, like
(3) Red is a colour
to propositions in formal mode, like
(3') 'Red' is a colour-word
which is a bipolar proposition about a physical object, namely the word
'red'. This move does not escape Wittgenstein's strictures, since the Tractatus
treats as formal concepts not just ontological but also logico-linguistic cate­
gories like 'name' and 'proposition'. But the moot question is whether these
335
 SCEPTICISM
strictures are justified. The real shortcoming with Carnap's suggestion is that
(3') only captures the necessary status which is essential to (3) if it is about a
symbol, a SIGN which means a particular object. But the ontological category
of that object, and hence the logico-syntactic category of the symbol, is once
more an internal property. It follows that (3') could no more be bipolar than
(3). _
Wittgenstein's own later solution is to abandon the idea that only empiri­
cal propositions are meaningful. The Tractatus's arguments show only that
propositions employing formal or 'logical concepts' (OC §§36-7) do not pro­
vide us with insights into the essence of reality, or with new information (a
point preserved in the idea that language is AUTONOMOUS). It does not follow
that such propositions are nonsensical pseudo-propositions (RFM 395-6,
402-3). Formal concepts have legitimate uses in GRAMMATICAL propositions,
as explanations of meaning or philosophical reminders. (3') and (3) can be
used to express the rule that whatever can be said to be red can also be
said to be coloured. One use of such grammatical propositions is to exclude
as nonsensical the sentences which generate paradoxes like Russell's, Grel-
ling's or the Liar. Like the Tractatus, the later Wittgenstein holds that these
paradoxes can be dispelled not through a consistency proof, but through an
analysis of the terms used in constructing them (WVC 121-4). His own ana­
lysis resembles Ryle's. Paradoxical sentences have no application in the lan­
guage-game, they resemble a game like 'thumb-catching' (RFM 120-3, 367;
LFM 206-9). This means that they make no statement, and hence cannot
be used to derive a contradiction.
(4) This statement is false
makes a statement only if 'this' refers to a form of words which make a true
or false statement. But if 'this' refers to (4) itself, the question of what state­
ment, if any, is made cannot be solved without vicious circularity. One can
imagine a use for paradoxical sentences in a logical exercise. But (4) cannot
be used to make a self-referential statement about which we can raise the
unanswerable question of whether it is true or false (RFM 404; RPP I §§65,
565; Z §691; see CONTRADICTION).
scepticism This is the view that knowledge is impossible, either in gen­
eral or with respect to a particular domain. Modern scepticism derives from
Descartes. It is based on the assumption that for a proposition to be known
it must either be evident, that is, self-evident or evident to the senses, or be
adequately supported by evident propositions. For the Cartesian and empiri­
cist traditions evident propositions are those about subjective appearances
('It seems to me just as if I were perceiving such and such'), which are sup­
posed to be immune to doubt. The sceptic challenges our right to pass from
such statements to propositions about mind-independent things ('I perceive
336
SCEPTICISM
such and such'). Various foundationalist responses try to meet this challange:
inductivism (inference to the best explanation), reductivism (idealist or phe-
nomenalist), transcendental philosophy and the defence of common sense
(Reid). There have also been indirect responses, which try to defuse the
sceptical challenge, by rejecting the very questions it poses. The 'scandal of
philosophy' is not that a proof of the existence of the external world is yet
to be given (Kant), but that 'such proofs are expected and attempted again
and again' (Heidegger).
One indirect response is Humean naturalism: our beliefs cannot be justi­
fied, because the sceptic's reasoning itself is perfectly legitimate and correct.
But due to our natural dispositions we cannot help holding the beliefs
attacked by scepticism, which hence need not be taken seriously outside the
realm of philosophy. Wittgenstein agreed that sceptical doubts cannot be
refuted in the sense of being shown to be false. But he insisted that scepti­
cism is flawed in a way which can be exposed by rational argument, namely
by virtue of being nonsensical. Against Russell's Humean stance that scepti­
cism is 'practically barren' though 'logically irrefutable', he remarked 'Scep­
ticism is not irrefutable, but obviously nonsensical, when it tries to raise
doubts where no questions can be asked. For doubt can exist only where a
question exists, a question only where an answer exists, and an answer only
where something can be said' (TLP 6.51; NB 1.5.15).
The work leading up to and including Philosophical Investigations provides
glimpses of such a critique of sense, a critique that helped to inspire the
anti-sceptical arguments of linguistic philosophy. 'If we are using the word
"to know" as it is normally used (and how else are we to use it?), then
other people very often know when I am in pain' (PI §246). According to
the rules of our grammar, it makes perfectiy good sense to say that I know
that others are in pain. This suggests that the sceptic is like someone who
claims that there are no physicians in Reading, since by 'physician' he
understands someone who can cure any disease within twenty minutes. His
doubts either amount to an ignoratio elenchi, since they employ 'knowledge'
according to other rules than the knowledge claims they purport to attack,
or express the sceptic's rejection of those rules (BB 55-61). But according
to the AUTONOMY OF LANGUAGE that rejection cannot be justified by reference
to the essence of reality. Nor could it be argued that our rules are pragma­
tically inferior to those implicit in the sceptic's position. Our concepts draw
important distinctions (e.g., between more or less well-established beliefs)
which he obliterates.
Both the Cartesian sceptic and his foundationalist opponent assume that
We at any rate know, incorrigibly, how things appear to us on the basis of
introspection. Wittgenstein's attack on the INNER/OUTER picture of the mind
turns the sceptic's picture on its head. We can know about the material
world, but not about the postulated mental realm: first-person present tense
337
SCEPTICISM
psychological statements are (typically) AVOWALS rather than descriptions of
an inner realm based on infallible introspection. Moreover, Wittgenstein
suggested that the language of subjective appearances is semantically para­
sitic on the language of perceptual objects and qualities. We learn it later,
and the sense of 'It is raining' is presupposed by that of 'It looks to me as if
it is raining.' The expression of what is subjectively seen is not a description
of private objects from which we precariously infer descriptions of public
objects, but a new linguistic technique, namely of making tentative judge­
ments about material objects (Z §§420-35). These ideas also undermine ego­
centric foundationalism (inductivism, reductivism). But they resemble Kant's
transcendental argument that the possibility of ascribing perceptual qualities
to mind-independent objects is a precondition of the possibility of ascribing
mental states to oneself, except that Wittgenstein would deny that the latter
are descriptions or cognitive claims.
Wittgenstein's most substantial discussion of scepticism is contained in On
Certainty. Its inspiration was Moore's defence of common sense. Moore
claimed that there are empirical truths which we can know with certainty -
for example, 'The earth has existed for a great many years.' Moreover, he
maintained that these common-sense truisms provide a rigorous proof of the
existence of the external world, since the premises are known for certain
and entail the conclusion. He held up his two hands and said 'Here is one
hand and here is another, so there exist at least two material things.' On Cer­
tainty conducts a three-cornered argument with Moore and the sceptic. Witt­
genstein grants that Moore is CERTAIN of these common-sense truisms, but
denies that he knows them. He also rejects Moore's claim to have proven the
philosophical proposition 'There are physical objects', since his common-
sense premise begs the question. For the sceptic, a doubt remains, namely
why should looking at my hands guarantee anything? After all, he is not
challenging a move within our established LANGUAGE-GAMES, for example that
Pluto exists. That sort of doubt can be resolved by observations and calcula­
tions. By denying that there are any ways of making sure, he is challenging
the whole language-game of physical-object discourse (OC §§19, 23, 83,
617). In claiming to know that he has two hands, Moore takes for granted
the conceptual framework which the sceptic attacks.
Wittgenstein tries to undermine both positions by impugning the sense of
the very proposition There are physical objects' (OC §§35-7, 57). It is not
an empirical proposition: on the sceptic's view, whether there are physical
objects makes no difference to the course of our experience, which is correct
in so far as we cannot even specify what it would be for there to be no
physical objects. Unlike, say, 'A chair is a physical object' it is not a gram­
matical proposition either, since it is not used to explain the meaning of
'chair' or 'physical object', and does not stipulate, for example, that one can
proceed from 'A chair is in the room' to 'A physical object is in the room.'
338
SCEPTICISM
\t best, like 'There are colours', it means that there is a category of words,
namely 'physical-object words'. This position is heir to the Tractatus: like the
formal concepts of the SAYING/SHOWING distinction, 'logical concepts' like that
of a physical object cannot be used in empirical propositions, but are evi­
dent from the logical behaviour of physical-object words. It is also close to
Carnap's proposal that the question of whether there are physical objects is
an external one which, unlike internal questions ('Are there dodos?'), boils
down to the question of whether to adopt a certain conceptual framework
(although Carnap's idea that we might instead opt for a sense-datum lan­
guage is incompatible with Wittgenstein's claim that the latter is parasitical
on our physical-object language). Both philosophers also hold that doubting
makes sense only if something could speak for or against the doubt, and
that hence a sceptical challenge like 'Things may change when unobserved
and change back when observed' is meaningless (OC §§117, 214-15; see
VERTFICATIONISM)
Scepticism and foundationalism alike ignore that doubt and the allaying
of doubt (justification) make sense only within a language-game. The lan­
guage-game itself can be neither justified nor doubted, is neither reasonable
nor unreasonable (OC §§559, 609-12). What kind of ground it makes sense
to require or adduce in favour of a claim is part of the meaning of that
claim, and hence subject to grammatical rules. These rules set limits to
meaningful doubt, by determining what could possibly count as questioning
or vindicating a claim of a particular kind. Doubt and justification make
sense only relative to the rules guiding the use of the expressions involved.
They come to an end when, after going through the ordinary procedures
for assessing a claim, we are confronted with doubts which are not provided
for by our rules, that is, which do not count as legitimate moves in the lan­
guage-game (OC §204; PG 96-7, 101). If I have justified a claim in the
ways licensed by these rules, I can only react to further challenges by reject­
ing them.
When challenged to show that the ripe tomato I look at in plain daylight
is red my only reply is that thisis* is simply what we call 'red'. If pressed
further I could only point out that this is how we speak, that is, reject the
challenge as meaningless. 'It is part of the grammar of the word "chair"
that this is what we call "to sit on a chair'" (BB 24; OC §§624-5; PI
§§380-1). Such claims were one source of the paradigm-case argument
employed by linguistic philosophers in the fifties: if this (pointing to the
chair) is what we call 'a chair', then in stating that it is a chair we could
not fail to state the truth. However, Wittgenstein insisted that 'This is a
chair' is exempt from doubt only as an OSTENSIVE DEFINITION which uses the
chair as a sample. In that case we have not refuted the sceptic by proving
'an indubitable truth, but have excluded his doubts as nonsensical through a
grammatical stipulation. Equally, if a sceptic about INDUCTION remonstrates
339
 SCEPTICISM
that it is only in the past that such and such a regularity has been
observed, he ignores that there is no such thing as now having evidence
from the future. What we call 'evidence that something will happen' is past
observations (OC §275).
Moore's common-sense truisms mark points at which doubt looses its
sense. They are the background against which we distinguish between true
and false, and therefore 'hinges' on which even our doubts turn (OC §§94,
341-3, 401-3, 514-15, 655). At least some of them are empirical in that
they state contingent facts, that is, their negation is not ruled out as non­
sensical by GRAMMAR. Nevertheless, the possibility of their being false is
restricted by the fact that not only our web of beliefs, but also our language-
games depend on them. If they should turn out to be false, other proposi­
tions would loose their sense. For we can distinguish between true and false
only against this backcloth. Consequentiy, we can call these propositions
into question only by doubt which calls itself into question, rather like cut­
ting off the branch on which one is sitting. As the scope of the sceptic's
doubt increases, its sense contracts. 'Doubt gradually loses its sense. This
language-game just is like that' (OC §56, see §§494, 498).
This strategy is reminiscent of the elenctic or transcendental arguments
envisaged by Aristotle and Strawson: the sceptic's doubts are incoherent,
since their making sense tacidy presupposes the conceptual framework which
they explicidy attack. It is conclusive when directed against the idea of uni­
versal doubt, or scepticism concerning the laws of logic. But Wittgenstein
extends it to sceptical attacks on empirical knowledge. The 'hypothesis' that
nothing around us exists is like the hypothesis that all our calculations might
be wrong, or- that all moves we make in playing chess might be wrong - it
removes the grounds for speaking about 'hypotheses', 'calculations' or 'play­
ing chess'. If one is asked to bring a book, and doubts that the thing over
there really is a book, one must either know what people mean by 'book' or
be able to look it up or ask someone - which itself presupposes knowledge
of what other words mean. But that a given word means what it does is
itself an empirical fact. Hence, to engage in doubt, some empirical facts
must be beyond doubt (OC §§55, 514-19).
Wittgenstein also applies this strategy to Descartes' dreaming-argument.
He claims that dream-reports are AVOWALS rather than descriptions (PI §448,
LI 184, 222-3; see LC 41-52 for a discussion of Freud's theory of dreams).
In On Certainty, he intimates that the, dreaming-argument ignores that one
cannot entertain occurrent thoughts while dreaming (OC §§675-6). This
argument, elaborated by Malcolm and Kenny, is better than its reputation.
Wittgenstein has a reasonable case for holding that the possibility of occur-
ent THOUGHTS is linked to the possibility of avowing these thoughts, and
hence incompatible with sleep (he would argue that although things can
occur to one during sleep, these are not beliefs one holds). Accordingly,
340
SCIENCE
whenever I entertain the question 'Am I awake?' I can answer it affirma­
tively, and without having to rely on any evidence. One might think that
even if the sceptic grants the difference between dreaming that p and think­
ing that p, he can challenge me to show that I actually believe that I am
awake, as opposed to merely dreaming it. But this ignores that I can only
be challenged, etc., if I am awake, otherwise I shall merely dream that I am
being challenged. 'The argument "I may be ch-eaming" is senseless for this
reason: if I am cfreaming, this remark is being dreamed as well - and
indeed it is also being dreamed that these words have any meaning' (OC
§§383, 642). To dream that a certain string of words makes sense does not
entail that it does make sense ('Da/)' never entails '/>'). To be sure, it does
not exclude the possibility of their making sense either, since not everything
that is dreamed is false. But Wittgenstein's point is that we cannot even
entertain a doubt whether we understand our own language, without stand­
ing 'before the abyss' (OC §§369-70, see §§114, 126), that is, without mean­
ingful discourse coming to an end.
It has been maintained that the sceptic could cheerfully accept that his
doubts violate preconditions of the possibility of language, since he rejects
the possibility of semantic knowledge as well. Cheerfully perhaps, but not
coherendy. A claim like 'I cannot know what these words mean' is self-refut­
ing: if it is true it must be meaningless. If Wittgenstein can drive the sceptic
into this corner, he has prevented him from making a coherent contribution
to the debate. That is not the same as refuting the sceptic, but it is not a
second-best: to silence a doubt by means of argument is to resolve the philo­
sophical problem.
science While Wittgenstein had an abiding interest in engineering and
certain kinds of scientific investigation, his cultural attitudes were inimical to
the scientific spirit of the twentieth century. But this ideological stance can
be separated from his methodological position. The latter rejects not science
but scientism, the imperialist tendencies of scientific thinking which result
from the idea that science is the measure of all things. Wittgenstein insists
that PHILOSOPHY cannot adopt the tasks and methods of science. His early
work was influenced by the Neo-Kantian philosopher-scientists Hertz and
Boltzmann. They reflected on the nature of science in order to free it from
metaphysical elements, sharply distinguished between its empirical and a
priori elements, and linked the latter to the nature of representation. Science
forms pictures or models (Bilder) of reality, whose logical consequences corre­
spond to the actual consequences of the situations depicted. Its theories are
not only determined by experience, but actively constructed within the fra­
mework of a 'form of representation'. Within limits set by logic, these forms
are subject only to pragmatic constraints — simplicity and explanatory power
(Mechanics Introd.).
341
SCIENCE
The Tractatus makes explicit this Kantian contrast between science, which
represents the world, and philosophy, which 'sets' the logical 'limits' to the
'sphere of natural science'. Science explores the accidental and consists of
the 'totality of true propositions' (TLP 4.1 Iff.). The more specific discussion
of the nature of scientific theory (TIP 6.3ff.) distinguishes the following
phenomena:
(a) Empirical generalizations are molecular propositions, truth-functions of
ELEMENTARY PROPOSITIONS. They describe objects, and their totality is an all-
inclusive description of the world.
(b) 'Laws of nature', by contrast, depict reality only indirecdy. Newtonian
mechanics, for example, describes all physical facts through differential
equations, and in terms of forces acting upon point-particles. Its natural laws
provide the 'building blocks' of empirical science, by determining a 'form of
description'. They lay down how scientific propositions can be derived from
'axioms', and hence what form specific generalizations and descriptions can
take. But they do not themselves describe particular point-masses. Natural
laws do not describe necessities in the world, since the only necessity is logi­
cal. Indeed, they do not even provide EXPLANATIONS of why things happen as
they do. In the absence of physical necessities, what happens in the world is
a matter of brute contingency; it can no more be explained by reference to
the operation of inviolable natural laws than by mvoking fate (TLP 6.341,
6.343ff., 6.37ff.).
(c) The principles of specific scientific systems like Newtonian mechanics
differ from the a priori principles of scientific theorizing in general, notably
the laws of causation, induction, least action and conservation, which are
themselves a mixed bag. The law of CAUSATION signifies the insistence that
any event must be explicable through a natural law of some kind; the law of
INDUCTION, by contrast, expresses an empirical proposition, namely that our
forms of description will continue to fit future facts in the way they have
done in the past (TLP 6.31-6.321, 6.36L, 6.362-6.372; RUL 1.14).
In formulating natural laws within the constraints of a chosen physical
theory, we proceed through the 'process of induction', which means that we
opt for the simplest law that can be reconciled with our experience. This law
is then employed as the basis for predictions, on the assumption of the
'principle of induction'. We assume that nature is simple and uniform — but
there can be no logical justification for this assumption (TLP 6.31, 6.363f).
Accordingly, laws of nature are rules for the derivation of predictions; and
the principles which underlie particular scientific theories are conventions.
There is only one LOGICAL SYNTAX. However, within its fimits, different scien-
342
SCIENCE
tific theories (Newtonian vs. relativistic mechanics) are guided by different
'systems' or 'forms of describing the world' (Formen der Weltbeschreibung). These
determine how empirical phenomena can be depicted within their frame­
work, and hence are not themselves accountable to experience. Wittgenstein
illustrates this through the analogy of describing irregular spots on a surface
with the aid of a 'network' (TLP 6.34If.; NB 6.12.14, 17.1./25.4./20.6.15).
On the one hand, any figure can be recorded to any degree of precision by
a sufficientiy fine mesh (if necessary, by moving the points of origin of the
grids); the shape of the mesh (e.g., square or triangular) is 'optional'; and the
use of a network brings the description into a 'unified form', which is given
a priori. On the other hand, it is a posteriori, and shows something about
reality, that a given figure can be described most simply by a net of a
specific shape and fineness.
This picture of science is conventionalist, in the vein of Hertz and Boltz-
mann. In spite of its cryptic style, and the scarcity of illustrations, it became
one of the major inspirations of instrumentalist conceptions of science.
Unlike the Tractatus, Ramsey and Schlick held that laws of nature are gen­
eralizations; but they tried to distinguish them from accidental general­
izations by treating them as rules rather than propositions. Wittgensteinian
instrumentalism improves on earlier versions in that it does not regard scien­
tific theories as premises of scientific predictions - which would mean that
they must be true or false, and hence descriptions - but rather as rules
which license scientific inferences. Nevertheless it remains open to serious
objections. For one thing, the denial that natural laws provide explanations
seems guided by a rationalist ideal of explanation, according to which A
explains B only if A logically entails B. For another, the fact that scientific
theories can be used to construct predictions does not entail that they are
not descriptions. Why not say that Newton's laws describe, or are proposi­
tions about, how bodies move in the absence of friction?
The later Wittgenstein would have accepted this reply, because he adop­
ted a more catholic conception of PROPOSITIONS (neither is there a trace of
rationalist prejudices concerning explanation). But he continues to insist that
scientific theories or laws of nature differ from straightforward descriptions
of particular objects - by virtue of the role they play within scientific belief
formation. His scattered remarks prefigure Kuhn (AWL 16, 39-40, 70—1,
98; BB 23, 56-7; RPP I §225; OC §§512-16): what the latter calls a scien­
tific 'paradigm', which informs the way a scientific theory responds to evi­
dence, Wittgenstein calls a FORM OF REPRESENTATION. For example, Newton's
first law of motion is not an empirical proposition which is up for grabs, but
a 'norm of representation', which guides the physicist's reaction to recalci­
trant evidence. If a body does not rest or move with constant motion along
a straight line, we postulate that some mass acts upon it; and if there are no
visible masses, we postulate 'invisible masses', as did Hertz. The introduction
343
SCIENCE
of a new form of representation (e.g., the Copernican revolution or Freud's
idea of 'unconscious desires') may result from an empirical discovery, but is
not itself a discovery forced upon us by facts. Rather, it is to adopt a new
'notation' for reasons of simplicity, explanatory power, etc.
Whether or not it is correct, this conventionalist account does not reduce
scientific revolutions to attaching old labels ('desire') to new things (as is
often claimed). A form of representation determines the meaning of key
scientific expressions. But it does more than simply label things; it provides a
way of making sense of experience, of making predictions and thus informs
complex scientific practices. Changes to our form of representation are far
from trivial as concerns their grounds and consequences: they result not in
mere re-naming, but in a new way of theorizing about the world. Indeed,
some elementary scientific propositions ('Water boils at 100° C) are so cen­
tral that although they can in principle be refuted by experience, this would
in effect be to 'change our whole way of looking at things' (OC §292, see
§§108, 293, 342, 599-608).
The later Wittgenstein's main aim is not to provide an account of science
itself, but to contrast it with PHILOSOPHY, AESTHETICS and psychoanalysis (PI
§§109, 126; AWL 37-40; LC 11-29). This contrast is independent of the
tenability of his conception of science, since it presupposes only that scien­
tific theories and hypotheses try to provide causal explanations of empirical
phenomena. Philosophical problems, by contrast, cannot be solved by
experience or causal explanation, since they are conceptual, not factual.
They require not new information or discoveries, but greater clarity about
GRAMMAR. This means that there should be a division of labour between
science and philosophy's second-order reflection on our conceptual appara­
tus. Alas, the twentieth century obsession with science makes it difficult to
uphold this division, and thereby obstructs philosophy (CV 16; PR Pref.; BB
17-18):
The scientific procedure of explaining diverse phenomena by reference to
a small number of fundamental laws induces a 'craving for generality'
and a 'contemptuousness for the particular case': we seek analytic defi­
nitions when we should be mapping the various uses of words.
Science tries to make phenomena intelligible through causal explanations,
while Wittgenstein thought that philosophical problems should be solved
through an OVERVIEW of phenomena in the spirit of Goethe and Spen­
gler (although occasionally he extends the idea of an overview to scien­
tific problems).
The scientific obsession with progress leads us to believe that philosophi­
cal achievement must lie in the construction of ever grander theories,
not in the clarification of concepts.
We are prone to believe that only science, especially physics, can tell us
344
SIGN/SYMBOL
what is real, and that, for example, secondary qualities are merely sub­
jective. Wittgenstein regards such claims on behalf of science as concep­
tual confusions which must be subjected to a philosophical critique.
In addition to this methodological resistance to scientism, Wittgenstein
also developed an ideological contempt for the 'idol worship' of science,
which he regarded as both a symptom and a cause of cultural decline (RW
112, 202-3; CV 6-7, 49, 56, 63). Partly, this reaction indicates his cultural
conservativism. However, it also expresses a humanistic worry that the pre­
dominance of science and the advance of technology and industrialization
marginalize ETHICS and art, and thus endanger the human spirit. Yet, even
while regretting the pernicious influence of the scientific spirit, Wittgenstein
distinguished between good and bad scientific works (RW 117; LE 4; LC
27-8; CV 42). The former follow ideals of clarity and intellectual honesty,
and involve detailed empirical investigations, like Faraday's Chemical History of
a Candle. The latter, like Jean's Mysterious Universe, pander to a craving for
mystery, and engage in speculation.
sense see MEANING
sign/symbol The Tractatus distinguishes between signs (Zekhen), percep­
tible sounds or inscriptions, and symbols, signs which have been projected
onto reality. A PROPOSITION is a 'propositional sign in its projective relation
to the world' (TLP 3.12), it has a sense because it has been correlated with
a situation; similarly a name is a sign which has meaning because it has
been correlated with an object. At one level this distinction between a mere
sign (sound or squiggle) and a significant sign or symbol is straightforward,
but it is linked to several complex issues.
(a) According to the Tractatus's initial explanation, a sign is 'what can be per­
ceived of a symbol'; a 'symbol' or 'expression' is a proposition or part of a
proposition which 'characterizes' or is 'essential to' the sense of the proposi­
tion, and can be shared by different propositions (TLP 3.3Iff., 3.32).
Accordingly, if there is something which different propositions all say, then
there is an expression which characterizes this class of propositions - for
example, '4 is red', '4 is green', etc. all say that 4 is coloured. This cannot
be expressed by a bipolar proposition, since it involves a formal concept like
'colour', but only through the use of a propositional variable whose values
are all propositions ascribing a colour to 4 (see SAYING/SHOWING).
(b) There are puzzles about the criteria of identity for symbols, (a) suggests
that all signs with the same logical function, that is, all names with the same
l
meaning and all propositional signs with the same sense (e.g., p V q' and
345
SIGN/SYMBOL
'~(~p.~q)') express the same symbol. That is why in an ideal notation, which
reveals 'the symbol in the sign' (TLP 3.325f.), every object will only have
one name and every name will name only one object (see IDENTITY), and all
propositions with the same truth-conditions will be expressed by the same
T / F symbol (see TRUTH-TABLES). This suggests that signs are tokens (inscrip­
tions or utterances), while symbols are types which are expressed through
these tokens. But that suggestion conflicts with two other points: (i) a symbol
is a sign in its projective relation to reality; (ii) the Tractatus distinguishes
between 'accidental' or arbitrary, and 'essential' features of symbols. The
latter constitute 'what signifies in a symbol', namely what all symbols that
fulfil the same logical function (and hence are substitutable according to the
rules of logical syntax) have in common (TLP 3.34ff.). For example, the 'real
name' is what is common to all symbols which signify the same object (TLP
3.3411). Wittgenstein deliberately changed a distinction between accidental
and essential features of signs into one concerning symbols (PT 3.24ff.). Pre­
sumably, his reason was that while every symbol is a sign, namely one pro­
jected onto reality, a 'mere sign' does not include its method of projection,
and hence lacks essential logical features. Accordingly, we have to distin­
guish mere signs, symbols, and real symbols. Thus, 'p V q' and '~(~p.~q)' are
two different symbols, with different methods of projection; the real symbol
is what they have in common in the T / F notation, namely '(TTTF)(/>,a)'.
(c) The idea of a symbol is linked to that of a 'mode of signification'. Tracta­
tus 3.321-3.323 states that a single sign may be 'common' or 'belong to' two
different symbols, in which case they 'signify in different ways' (thereby
creating philosophical confusions). It has been argued that this idea of a
'mode of signification' (Bezeichnungsweise) corresponds to Frege's idea of sense,
a mode of presentation of a meaning. Tractatus 3.317 suggests that a mean­
ing (Bedeutung) is what is signified by a symbol; 3.3411 that a single object
may be signified by different symbols. If one assumes that these symbols
differ in their mode of representing the object, it seems to follow that unlike
their Russellian counterparts, Tractarian NAMES are correlated with objects
not direcdy, but via a Fregean sense.
Against this speaks the fact that Tractatus 3.321 states only that any differ­
ence in mode of signification is always a difference in symbol, not that any
difference in symbol is always a difference in mode of signification. Follow­
ing the idea that symbols are projected signs, different symbols referring to
the same object may differ not in their mode of signification, but simply in
being different signs. But even if one accepts instead the idea that symbols
are types, it is clear that there cannot be two names which refer to the same
object through different modes of signification, simply because in an ideal
notation there will be only one name for each object, which means that
modes of signification do not play a role in the logical functioning of a
346
SIGN/SYMBOL
name. Moreover, words signifying in different ways is a matter not of names
referring to a single object in different ways, but of words belonging to dif­
ferent logical categories, as 3.323 makes clear: it applies not to logically
proper names 'a' and 'b\ but to 'is' in its capacity as copula, identity-sign,
and expression of existence. It has been objected that charity demands that
3.3411 be interpreted as allowing that co-referential names can have differ­
ent senses (modes of signification), since otherwise it boils down to the trivial
point that we can use different signs to refer to the same object. In fact, this
interpretation is extremely uncharitable, since it makes Wittgenstein's discus­
sion of symbols contradict his explicit insistence that 'Only propositions have
sense' (TLP 3.3) and that names are directly correlated with their meanings
(TLP 3.221) and hence immune to referential failure.
(d) In the notation of the Tractatus, 'aRb' is a (propositional) symbol (it
depicts a's standing-in-relation-/? to b), while 'x-0' is a mere sign. It could be
turned into a symbol, by laying down a METHOD OF PROJECTION for its ele­
ments, that is, correlating them with elements of reality. This is somediing
human beings do - and only human beings, since it involves a process of
diinking: the noises made by a parrot could never be anything more than
mere signs (see THOUGHT/THINKING). In this respect, signs are conventional, as
the existence of different languages and notations shows. It is 'arbitrary'
what signs we use as symbols, and what differences in signs mark a differ­
ence in what is symbolized (TLP 3.322). But the rules of symbolism are not
arbitrary. Once we have laid down that a certain sign is to stand for a cer­
tain object, the combinatorial rules of the former are determined by the
LOGICAL FORM of the latter. The possibility of adopting a certain 'notation', of
projecting a system of signs in a coherent manner, 'discloses something
about the essence of the world' (TLP 3.342f.). It is arbitrary whether we
express negation through the presence or absence of '"', but not that any
sign expressing negation must reverse the sense of what it applies to. It does
not matter whether we use the jV-operator or the standard set of LOGICAL
CONSTANTS, but it does matter that the latter are replaceable by the former.
It does not matter whether we call a complex 'Tully' or 'Cicero', but both
signs must be analysable in the same way into ELEMENTARY PROPOSITIONS con­
sisting of logically proper names.
The later Wittgenstein does not use the Tractatus's complex contrast
between signs and symbols. He continues to discuss the question of what
gives MEANING to signs, but in a way which diverges from the Tractatus. The
latter acknowledged that in order to 'recognize the symbol in the sign', that
is, in order to recognize its logical role, one must pay attention to its 'sig­
nificant use' (TLP 3.326). But what makes for the significant use is a mental
process of thinking which accompanies the use of signs. By contrast, the
347
 SOLIPSISM
later Wittgenstein rejects the search for 'the real sign' in our words (PI §105;
OC §601). What gives life to signs is not an entity or a process associated
with them, but their USE.
solipsism This is the view that nothing exists apart from oneself and the
contents of one's mind. Although this idea has rarely been endorsed expli-
citly, idealists or phenomenalists have been tempted by or even implicitly
committed to it. The discussion of solipsism (TLP 5.6-5.641) marks the
intersection of the logical and mystical parts of the Tractatus. The 'key to the
problem, how much truth there is in solipsism' is that 'the limits of my language
mean the limits of my world.' What the solipsist means is that 'the world is
my world.' This inexpressible truth manifests itself in 'the fact that the limits
of language (of the only language which I understand) mean the limits of my
world' (TLP 5.62 refers to 5.6, see NB 23.5.15).
Russell had linked solipsism not just to the limits of knowledge, but also
to those of language. According to his principle of acquaintance, every
meaningful word must stand for something within the individual's immedi­
ate present experience. This suggests a semantic 'solipsism of the present
moment' according to which only the sense-data I am presendy aware of
are real. Russell escapes this conclusion by an inductive inference to the
conclusion that there probably are other minds (Logic 130-4; Problems 8-9).
Wittgenstein repudiated this approach to SCEPTICISM, but developed the
linguistic perspective on solipsism. His main source, however, was transcen­
dental idealism. Kant had refuted the Cartesian doctrine of a soul-substance,
but introduced two other notions: the 'transcendental unity of apperception',
a formal feature of judgements, namely that they can be prefixed by 'I
think'; and a 'noumenal self, the locus of free will and the moral law. Scho­
penhauer elaborated the former notion by claiming that the 'subject of
knowledge' to which the world as representation appears is merely an 'in­
divisible point'. It cannot be encountered in experience, just as the eye 'sees
everything except itself. Nevertheless, it is a 'centre of all existence' and
determines the limits of the world. For 'the world is my representation', and
the idea of a world without a representing subject is a contradiction in
terms (World I 3-5, 15, 332, II 277-8, 491). Schopenhauer replaced the
noumenal self by a superindividual cosmic will which underlies the world as
representation. I know my body as the embodiment of this will, because I
am direcdy aware of my actions. As regards both cognition and volition, the
individual, the 'microcosm', is identical with the 'macrocosm' (World I 103-
6, 162, II 486 - an idea taken up by Weininger). Schopenhauer disavows
the solipsistic implications of this identification. But, like Russell, he grants
that solipsism is irrefutable, and departs from it only through insisting that
the subject of experience is not a mental substance, and that everything is a
manifestation of the superindividual will.
348
SOLIPSISM
The early Wittgenstein disregarded this facile disclaimer. He developed a
transcendental solipsism through metaphors and topics (MYSTICISM, ETHICS,
WILL) derived from Schopenhauer. Russell had insisted that acquaintance
requires a subject, with which we are acquainted, or which we know by
description. Wittgenstein, by contrast, rejects the idea of the 'thinking, pre­
senting subject'. In a book entided The World as I found it no self would
(Hume) or could (Schopenhauer) be mentioned. Like the eye of the visual
field, the self is not a possible object of experience; and it cannot be inferred
from the content of experience either. There is a 'human soul' which is the
legitimate subject-matter of psychology, but it is not a unitary self or subject,
but only an array of mental episodes (TLP 5.631-5.641; NB 7.8./11.8.16; cp.
Problems 27-8; Mysticism ch. X; 'Theory' 36-7; Logic 125-74).
Like Kant and Schopenhauer, Wittgenstein combines this rejection of the
Cartesian soul with accepting a 'metaphysical subject' or 'philosophical I'
which enters philosophy through the fact that 'the world is my world.' This
metaphysical subject is not a part of the world, but is nevertheless its
'centre', being both 'a presupposition of its existence' and its 'limit'. The
relation of what we experience, our field of consciousness, to the subject of
experience is analogous to that of the visual field to the eye — not the sense-
organ, but what he later called the 'geometrical eye'. This self is an 'exten-
sionless point', and the human individual a 'microcosm' (NB 11.6./4.8./
12.8./2.9./12.10.16; TLP 5.63, 5.633-5.64; LPE 297-9; BB 63-5). Argu­
ably, this metaphysical subject is identical with the 'willing subject', which is
the bearer of good and evil (TLP 5.633, 5.641; NB 21./24./29.7./2.8./
2.9.16).
The prima facie case for detecting some version of solipsism in the Tractatus
is overwhelming. Wittgenstein concedes not only that solipsism contains a
kernel of truth, as 'anti-solipsist' interpreters have it, but that 'what the solip­
sist means is quite correct' - namely that the world is my world. The only
criticism is that the solipsist tries to say what can only be shown, which is
the predicament of the whole Tractatus (see SAYING/SHOWING). Moreover, Witt­
genstein writes in propria persona about the metaphysical self, that 'world and
life are one' and 'that I am identical with my world'. Finally, the Notebooks
are full of purple passages which identify the world with life, life with con­
sciousness in general, and consciousness with the metaphysical self, and
judge philosophical views by their compatibility with the 'stricdy solipsistic
point of view' (TLP 5.621f.; GT 8.12.14; see NB 11.6./1.-2.8.16).
On the other hand, Wittgenstein claims to have travelled from 'idealism'
through 'solipsism' to 'pure realism', because 'I too belong with the rest
of the world.' Wittgenstein attacks Schopenhauer's view that while the rest of
the world is mere representation, the human body is a direct embodiment of
the will. Moreover, in contrast to the self of traditional solipsism, both the
willing and the metaphysical subject seem impersonal, a 'world soul' stripped
349
SOLIPSISM
of all individuality. However, Wittgenstein wavers on this point: the 'ethical
will' is linked to individuals, and 'in a higher sense' the Schopenhauerian
'world-will' is 'my will', just as my representations are the world (NB 2.9./
12.-17.10./4.11.16; T I P 5.64f). Moreover, the 'pure realism' into which
solipsism collapses is compatible with an austere version of transcendental
solipsism in which the analogy of the visual field takes the place of the
transcendental unity of apperception. Although the subject of experience
cannot be part of experience, it is a logical feature of my experiences that
they belong to me. 'The subject — we want to say - does not drop out of
the experience but is so much involved in it that it cannot be described'
(PG 156). Any representation of the world occurs from a perspective which
is uniquely mine. Because representation is linguistic, transcendental solip­
sism takes a linguistic turn. The 'connection between solipsism' and 'the way
a sentence signifies' is that 'the I is replaced by the sentence and the rela­
tion between the I and reality is replaced by the relation between the sen­
tence and reality' (BT 499). And that relation depends on the metaphysical
subject, a linguistic soul which breathes life into mere signs:
(a) Language is my language, because mere SIGNS turn into symbols through
my 'thinking the sense of the proposition' (TLP 3.11).
(b) The method of projection which underlies this language of thought
finks names with objects through acts of meaning (see OSTENSIVE DEFINI­
TION).
(c) These acts are arguably performed by the will of the metaphysical self:
'Things acquire "meaning" only in relation to my will'; and a Schopen­
hauerian will is invoked to explain intentionality (NB 15.10.16; PG
144-56).
(d) I can correlate with names only objects I experience, and what I can­
not project is not language. 'I have to judge the world, to measure
things' (NB 2.9.16), namely by injecting contents into logical forms.
Transcendental solipsism is compatible with empirical realism: it does not
assert that 'I am the only person that exists' or reject empirical propositions
about the external world or 'other minds'. The truth of solipsism manifests
itself in the very possibility of representation, and more specifically, the logi­
cal form of all empirical propositions: fully analysed 'A is in pain' refers only
to pain-behaviour of which / am aware, while 'I am in pain' directly refers
to my experience.
In his later work, Wittgenstein's perspective is 'the diametrical opposite of
solipsism', and he compares the solipsist to a fly in a fly-bottle (LPE 282,
300; PI §309). His first attempt to find a way out was the methodological
solipsism of his VERIFICATIONIST phase, which he shared with Carnap and
Schlick (PR ch. VI; WVC 49-50; M 100-3). He abandons the metaphysical
350
SOLIPSISM
subject of transcendental solipsism, but retains the idea of a mono-centred
language. First-person present tense propositions referring to immediate
experiences constitute the basis of language. Although such a language can
have anyone as its centre, a language with me as its centre is particularly
adequate. Third-person psychological propositions such as
(1) A feels pain
are analysed by reference to me, namely as
(1') A behaves as I do when I'm in pain.
Curiously, however, the T here does not signify an ego which owns these
experiences, but only an ineffable centre of language. In
(2) I feel pain
T is redundant, because 'pain' is defined as something / have, and no one
else could have (see I/SELF; PRIVACY). This position makes it difficult to draw a
contrast between A's displaying pain-behaviour, and A's being in pain;
BEHAVIOURISM is the flipside of solipsism. Moreover, private experiences enter
into the verification of (2), but not of (1). Because of this difference in
verification, mental terms do not mean the same in first- and third-person
utterances.
This exotic conclusion follows not just from methodological solipsism, but
from any approach to the problem of other minds based on the INNER/OUTER
picture. That picture accepts that statements like (1) are conjectures based
on behavioural evidence, while those like (2) are infallible, since they refer to
our own private experiences. It attempts to resist the sceptical conclusion
that we can never know that there are other minds through the argument
from analogy: I infer that when other people behave in the way I do when
I am in pain, they are also in pain. But if 'one has to imagine someone
else's pain on the model of one's own, this is none too easy a thing to do;
for I have to imagine pain which I do not feel on the model of pain which I
do feel' (PI §302, see §398; BB 46). If a given psychological term means
THIS, which I have and no one else could conceivably have, then the belief
that there are other subjects of experience is not just uncertain, as the
sceptic has it, but rather makes no sense.
If this is correct, the 'semi-solipsism' of the inner/outer picture ultimately
collapses into the solipsism of the present moment which Wittgenstein dis­
cussed in the early thirties, which claims, firstly, that whenever anything is
realty perceived it is always I who perceive it, and it is this, my present
experience, that is perceived; and secondly, that I am the 'centre of the
world' and 'the vessel of life' in that the only reality is my present experi­
ence (LPE 299; BB 61-5). The solipsist purports to have discovered that the
world is realty identical with his experiences, when all he has done is to
351
STATE OF AFFAIRS
recommend a new form of expression, namely one in which attributions of
'real' experiences to anybody other than the solipsist are meaningless, and in
which we might say, for example, "There is a real toothache' instead of
'N.N. (the solipsist) has a toothache' (BB 58-9; PI §§401-2). That way of
speaking has no practical advantage, since it is not supposed to make a dif­
ference to the solipsist's behaviour. It cannot have a philosophical advantage
either since it cannot be justified by reference to an 'essence of reality' (see
AUTONOMY OF LANGUAGE).
Wittgenstein's other arguments against solipsism, which are not included
in Philosophical Investigations, have a Kantian flavour. The solipsist invokes our
ordinary concepts, but without the contrasts (e.g., between one's own pres­
ent experiences and what they are experiences of) which are essential to
these concepts. Thus, not only does he oscillate between different gramma­
tical systems, but also, without these contrasts, terms like 'mine' or 'present'
are vacuous, (a) In the solipsist's mouth the term 'present' has no 'neigh­
bours'; it does not stand in contrast with past or future, and is hence redun­
dant. Equally, 'my pain' can be used to make a significant claim only in a
system in which it contrasts, for example, with 'her pain', not in one in
which it reduces to 'there is pain' (PR 84-6; WVC 50, 107; M 100-3; LPE
297; BB 71-2). The sense of the solipsist's proclamation 'Only my pain is
real' presupposes the grammatical system he purports to reject, (b) The first-
person pronoun T does not refer to, or trace, a continuous immaterial sub­
stance since the purported referent could constantly change without its use
being affected at all. Nor can it express the solipsist's allegedly unique per­
spective, since what he perceives simply is what is perceived. It merely indicates
the formal feature - the transcendental unity of apperception - that proposi­
tions can be prefixed by 'I think that' (LPE 283, 298-300; BB 60-9).
Unique ownership loses its sense with the disappearance of a unique self: to
be had by me is simply to be an experience. These arguments rule out any­
thing but a methodological, ego-less solipsism which insists that language
derives its meaning through private ostension to private experiences. By
demonstrating the incoherence of that view, the PRIVATE LANGUAGE ARGUMENT
pulls the carpet from underneath all forms of solipsism and idealism.
state of affairs see FACT
surveyability see MATHEMATICAL PROOF
symbol see SIGN/SYMBOL
352
T
tautology 'Tautology' is a Greek term which means 'repetition of what
has been said'. Its use in logic goes back at least to Kant (Logik §§36-7).
Kant characterized formal logic as analytic, but distinguished between two
types of analytic propositions, those in which the containment of the pred­
icate in the subject-concept is implicit, as in 'All bodies are extended', and
those in which it is explicit, as in 'All extended things are extended.' The
latter he labels 'tautological', and insists that unlike the former they are 'vir­
tually empty or devoid of consequences', since they do not even explicate
the subject. This corresponds to Leibniz's distinction between 'necessary'
and 'identical' truths, and to the contemporary distinction between analy­
tical truths proper and truths logical, the former being propositions which
can be reduced to logical truths through the use of definitions. In the nine­
teenth century, 'tautological' was used pejoratively to indicate that formal
logic, in particular the law of identity 'a = a', is trivial and poindess, since it
does not extend our knowledge. Wittgenstein would have encountered the
term in Coffey (see RCL) and Mauthner, who claimed that not just logical
and mathematical, but even empirical truths are tautological once known
(Beitrdge III 301, 324-5). And although Russell passionately denied that logi­
cal truths are tautological or purely analytic, Principia Mathematica (*1.2) labels
'(p v p) Z) p' the 'principle of tautology'. Even Frege admitted that a logical
truth like '/> 3 p' seems 'almost without content' ('Compound' 50).
Although the early Wittgenstein was not the first to characterize logic as
tautological, he was the first to use the term in a way which is both precise
and general, that is, not confined to either the principle of identity or pro­
positions involving literal repetitions. Moreover, he used it to distinguish dif­
ferent types of propositions that had been treated indiscriminately as
belonging to LOGIC. And he made out a convincing case for the idea that
logical propositions do not describe reality, but reflect linguistic rules.
According to Frege, the truths of logic are analytic in the sense of being
deducible from definitions and self-evident axioms. However, the axioms,
and hence indirectly the theorems, are characterized as truths which unfold
timeless relations between entities (thoughts and truth-values) inhabiting a
'third realm' beyond space and time (Foundations Introd., §§3, 26; Laws I
353
TAUTOLOGY
Pref.; 'Thought'). Russell's position was more Aristotelian than Platonist. He
treated logical truths as descriptions of the most general and pervasive fea­
tures of reality, and insisted that they contain only logical constants and
unbound variables. Thus, 'Whatever x, a and jS may be, if all as are /?s, and
x is an a, then x is a /?' is a logical proposition, but not 'If all men are
mortal, and Socrates is a man, then Socrates is mortal' (Principles 11; Principia
93; 'Theory' 98-101; External 66). But he and Frege share the conviction
that logic makes statements about entities or forms of some kind, just as the
empirical sciences make statements about physical objects.
Rejecting this assumption is the starting-point of Wittgenstein's philosophy
of logic. 'Logic must turn out to be of a TOTALLY different kind than any
other science.' The first step in delivering this promise is realizing that the
propositions of logic 'contain ONLY APPARENT variables', and that there
are no LOGICAL CONSTANTS (RUL 22.6.12, 22.7.13; NL 107; TLP 6.112). The
latter claim is directed against the idea that the terms of logic - proposi­
tional connectives and quantifiers — are names of entities, the former against
the ensuing idea that the propositions of logic are statements about these
entities.
Both Frege and Russell expressed the universal character of logical truths
through the use of 'real' variables, that is, variables which unlike 'apparent'
variables are not bound by quantifiers (Notation §§1, 14; Laws I §§17-18;
l
Principia ch. I). By this token, p V ~p' and \x\fx 3fa' are implicit general­
izations over propositions, concepts and objects. Initially, Wittgenstein
agreed that logical propositions are general, but he insisted that unlike
empirical generalizations they are essentially rather than accidentally true,
and hence cannot be expressed through signs containing real variables
(RUL 11.-12.13; NL 100). Unlike
(1) It is either raining or not raining
signs like 'p v ~p' are not themselves meaningful propositions, but only
'schemas of propositions' which use sentence-letters as dummies to indicate
the logical form of those propositions which are created by replacing the
dummies by meaningful sentences (this is close to Quine's account of the
role of sentence-letters in such schemas). Consequentiy, the complete gen­
erality of the law of excluded middle can only be expressed with the help of
quantifiers, namely as
(2) mv-p)
and logical propositions are generalizations of tautologies like (1).
'Notes dictated to Moore' substantially modifies this position, rejecting
propositions like (2) as nonsensical. Three reasons are implicit in Wittgen­
stein's new account. First, by quantifying over propositions, (2) assimilates
propositions to NAMES standing for objects, contrary to the sharp contrast
354
TAUTOLOGY
Wittgenstein elsewhere draws between the two. Second, the emerging
SAYING/SHOWING distinction puts on the index both (2) and its non-formal
equivalent 'Every proposition is either true or false', since they employ a
formal concept to characterize essential properties of propositions. Third
and most important, being general is neither necessary nor sufficient for
being a logical truth. Pace Russell, an ungeneralized proposition like (1) is
essentially true, and hence part of logic. On the other hand, general princi­
ples like the axioms of reducibility and of infinity or the law of induction are
contingent and have no place in logic (NM 108-9; TLP 5.535, 6.123 Iff.,
6.31). The propositions of logic are not generalizations of tautologies, but
themselves tautologies, that is, represented by schemas like 'p V ~p'.
Wittgenstein gives 'tautology' a precise meaning through the idea of truth-
functional composition. The truth-value of a molecular proposition depends
on those of the ELEMENTARY PROPOSITIONS of which it is a truth-function.
Among the truth-functional combinations of propositions there are two
'limiting cases'. Tautologies are combinations which are true (contradictions
like 'p.~p' false) no matter what the facts are, and this is displayed by their
TRUTH-TABLE taking a T (or F) for all truth-possibilities (assignments of truth-
values). Ordinary truth-functions have sense, since they depict possible states
of affairs truly or falsely. By contrast, tautologies and contradictions say
nothing. They do not delimit the world in any way, since the former are
compatible with all possible situations, the latter with none. They are not
'nonsensical', since they are licit combinations of propositions, but 'senseless'
in a quantitative way, that is, have zero sense. Unlike 'It is raining', (1) says
nothing about the weather. Tautologies 'give no information': 'If fifteen,
1
then fifteen! is no more an answer to the question 'How many people will
be present?' than 'Take it or leave it' is an order (TLP 4.46ff., 5.101; NM
118; BB 161; RFM 231; LFM 280).
Frege resisted the idea that logical truths are vacuous by pointing out that
they are 'undeniably true'. Wittgenstein would grant that they are 'on the
side of truth', but insist that they are 'degenerate propositions' in the sense
in which a point is a degenerate conic section. For they are made so as to be
true, since they combine BIPOLAR propositions in such a way that all factual
information cancels out, and it is for this reason that they can be neither
confirmed nor refuted by experience (RFM 167; see LFM 177-8; TLP
4.461, 4.465f, 6.121; NB 3729.10.14, 6.6.15; 'Compound' 50). Again; it
has been objected that at least complex logical propositions are far from
vacuous. But the crucial point is that while the truth of bipolar propositions
can be determined only by comparing them to reality, even complex tautol­
ogies can be recognized to be true 'from the symbol alone', namely through
calculations which use 'only rules that deal with signs' (TLP 6.113, 6.126).
Logical propositions are not truths about an ultimate reality, nor do they
express a special type of knowledge, as had traditionally been assumed; for
355
 TAUTOLOGY
they differ from all other propositions by virtue of being vacuous (TLP
5.1362, 6.111). Wittgenstein's account also casts more specific doubts on
Frege's and Russell's axiomatic representation of logic. There are no privi­
leged logical propositions ('axioms' or 'basic laws') from which all others
('theorems') are derived. It is a matter of indifference from which logical
proposition one commences; they are all of equal status, namely tautologies,
and they all say the same, namely nothing (TLP 5.43, 6.127f.; NB 10.6.15).
l
The T / F notation reveals that 'p V ~p\ ~(p.~p)', 'pDp', etc., are merely
different ways of expressing one and the same tautology, namely (TT)(p).
For any number n of elementary propositions, there are only two limiting
cases of truth-functional combination, and these are represented respectively
by formulae with 2" T's and 2" F's. Moreover, in an ideal notation all pro­
positions are expressed as truth-functions of the entire set of elementary pro­
positions. This threatens to collapse logic into a single immense tautology
(hence Sheffer's remark, 'There is but one Tautology and Wittgenstein is its
prophet' - WAM 86). Finally, to Moore the idea that all tautologies say the
same appeared as a reductio of the claim that they say nothing (M 61-6).
Wittgenstein tries to avoid both problems by maintaining that although all
tautologies say the same, they show different things about the logical proper­
ties of their constituent propositions (NM 114-17; TLP 6.12ff). Thus,
'~(P-~PY shows that '/>' and '~p' contradict each other, \(pDq)-p)Dq'
shows that 'q' follows from 'p Z) q' and '/>'. He thereby also provides room
for LOGICAL INFERENCE: being vacuous, tautologies cannot state that one pro­
position follows from others, but that '((P D q\. p) D q' is tautological (or
'((/> Dq)-P) .~q' a contradiction) provides 'the form of a proof (modus ponens).
The Tractatus contends that the only logically necessary propositions are
analytical, that is, vacuous tautologies (TLP 6. If., 6.3, 6.375). All proposi­
tions which seem to be true in whatever the circumstances without fitting
that bill must, on analysis, turn out to be either (a) empirical after all; (b)
tautological truth-functions; (c) nonsensical, attempts to say what can only be
shown, (c) holds not only of metaphysical propositions, but also of MATHE­
MATICAL equations. Like tautologies, they do not express a thought, but
unlike tautologies they are pseudo-propositions (TLP 6.2f.). (b) applies to
logical truths of the predicate calculus, for example \x)fxZ)fa', because 'fa'
expresses one of the possibilities which make up the possibly infinite con­
junction abbreviated by '(x)fx' - fa.fb.fc, etc' It also applies to '~(A is
red. A is green)', which allegedly cart be analysed into a truth-functional
tautology (TLP 6.1201-6.1203, 6.3751). The treatment of the former case is
undermined by Wittgenstein's realization that GENERALITY cannot be
explained in terms of infinite logical products, that of the latter by his sub­
sequent discussion of coLOUR-exclusion.
For a while, Wittgenstein tried to retain the idea that such propositions
are tautological by introducing into the truth-table notation rules which
356
THOUGHT/THINKING
exclude certain truth-possibilities, such as 'TT' for '2 is red' and 'A is green'
(RLF 34-7; WVC 73^1, 91-2). However, this amounts to abandoning the
idea that all necessity is analytical, based on the truth-functional combination
of logically independent propositions. It has been maintained that in
response to this difficulty Wittgenstein extended the notion of tautology to
all necessary truths. In fact, Wittgenstein continues to reserve that label for
the propositions of logic (BB 71; LPE 283; NPL 449; LFM 272-85). How­
ever, other necessary propositions resemble them in one respect. They are
not compared with reality, and hence are not descriptions of anything, and
a fortiori not of logical entities, but are to be explained by reference to lin­
guistic rules, 'p v ~p' is a vacuous tautology; but that it is a tautology gives
rise to a rule of inference which is neither ineffable nor tautological, but
part of our FORM OF REPRESENTATION in that it specifies how empirical propo­
sitions can be transformed (AWL 137^10; LFM 277-80; RFM 123, 231,
245-7; WVC 35, 106, 158-9; PR 125-30). Wittgenstein also abandoned the
idea that the fact that certain truth-functional combinations of elementary
propositions are tautologies shows the essence of the world (TLP 6.124,
6.13; NM 108-11; see BIPOLARITY).
Due to the Tractatus, the truths of the propositional calculus are com­
monly characterized as tautologies. The claim that logical propositions are
vacuous was accepted reluctantly by Russell, and enthusiastically by Ramsey
and by the logical positivists. The logical positivists used it against Kant's
idea that some a priori truths are synthetic. But they ignored Wittgenstein's
distinction between tautologies and mathematical equations, and his mythol­
ogy of symbolism. For them tautologies are consequences of arbitrary con­
ventions (the truth-tabular definitions of logical constants).
mought/thinking In the mentalist tradition, thoughts (cogitations, ideas)
were understood as psychic entities or occurrences which inhabit the minds
of individuals. In reaction, the anti-psychologistic and anti-idealistic move­
ment (Frege, Moore, Russell) reverted to a Platonic picture. Thus, Frege dis­
tinguished between private ideas (Vorstellungen), and thoughts, which are
abstract entities inhabiting a Platonic third realm. His grounds were (a) a
thought, that is, what someone thinks, is true or false independendy of
someone's thinking it; (b) two people can have the same thought; (c)
thoughts can be communicated ('Sense' 29-32; 'Thought').
Wittgenstein's early position seems to eschew both mentalism and Platon­
ism, but the issue is obscured by the fact that he uses 'thought' (Gedanke) in
two different roles. In its primary, Fregean use it signifies a proposition
(Satz). A thought is a 'logical picture of facts', that is, an optimally abstract
picture whose only pictorial form is its LOGICAL FORM and which does not
rely on any specific medium of representation. 'In a proposition a thought
finds an expression that can be perceived by the senses.' However, a
357
 THOUGHT/THINKING
thought is neither an abstract nor a mental entity correlated with a sen­
tence. Rather, it is an 'applied, thought (gedachte), propositional sign', a 'pro­
position with a sense' (TLP 3, 3.1, 3.5, 4). This means that a thought is a
sentence-in-use, a propositional sign which has been projected onto reality.
At the same time, the 'method of projection' which projects the proposi­
tional sign onto a state of affairs is 'to think the sense of the proposition'
(TLP 3.11; PT 3.12f.). In its second use, Gedanke signifies a mental entity
which has 'psychical constituents' that stand in the same sort of relation to
reality as the words which are the constituents of the propositional sign
(RUL 19.8.19). This suggests that a thought is a psychic fact which is not
identical but isomorphic with the propositional sign on the one hand, and
the depicted state of affairs on the other.
Perhaps Wittgenstein failed to notice the inconsistency because he held
that the mental process of 'thinking is a kind of language' (NB 12.9.16). A
thought is itself a proposition in the language of thought, and intimately tied
to the propositional sign. Just as a propositional sign is a significant proposi­
tion only if it is projected by a thought onto the world, so a relation
between psychical elements is a thought (rather than, say, a headache) only
if it is a projection of a propositional sign. Accordingly, it is essential to
thoughts that they can be completely expressed in language. This breaks
with the venerable view, shared by Frege and Russell, that the relationship
between thought and language is external. Thoughts are not entities beyond
language, and language is not merely a medium for ttanstnitting a pre-
linguistic process of thinking. At the same time, the Tractatus holds that the
LOGICAL ANALYSIS of a proposition of sign-language (Zekhensprache) will reveal
the structure of the underlying proposition in the language of thought.
Moreover, it remains wedded to the doctrine that it is the mind which gives
meaning to language by breathing life into sounds and inscriptions that
would otherwise be 'dead' (BB 3-5). While the precise nature of thinking is
relegated to empirical psychology, the production of thoughts is conceived of
as a process which must accompany speaking, and distinguishes it from the
squawkings of a parrot.
Wittgenstein later contended that the idea of a language of thought faces
a dilemma. On the one hand, thought must be intrinsically representational:
while my words can be interpreted by reference to what I think, my inter­
preting my own thoughts (save in the sense of asking myself why I have a
particular thought) makes no sense; unlike speech, 'thought is the last inter­
pretation' (BB 34-5; PG 144-5). On the other hand, this means that the
psychic elements do not stand in the same sort of relation to reality as
words. More generally, Wittgenstein criticized the view that thinking is a
mental process which accompanies speech and endows it with meaning (BT
ch. 6; PG ch.V; PI §§316-62). If thoughts are to give meaning to sentences
they must themselves have symbolic content. Yet this leads to a vicious
358
THOUGHT/THINKING
regress (see METHOD OF PROJECTION). This is obvious if one replaces the mental
accompaniment by a physical one: a sentence plus a painting is no less
capable of different interpretations than the sign by itself. To suppose that
the mind 'could do much more in these matters' because of its occult quali­
ties is a mythology of psychology (PG 99; Z §211).
Wittgenstein's second argument against the accompaniment conception of
thinking is that what distinguishes speaking with UNDERSTANDING from the
mechanical utterances of a machine or parrot is not an accompanying pro­
cess (he often - PI §§330-2, 341 - speaks of the difference between 'think­
ing speech' (denkendes Sprechen) and 'thoughdess speech' (gedankenloses Sprechen),
which wrongly suggests that he is concerned with the contrast of well-
considered and careless utterances, although that topic is broached in RPP
II §§250-67). Firsdy, such processes are neither necessary nor sufficient for
meaningful speech. Secondly, one cannot subtract the linguistic expression
to distil a separate process of thinking. Speaking with understanding or
thought is not like singing and accompanying it by playing the piano, but
like 'singing with expression' (PI §332). The difference lies in how it is done,
and in what the speaker is capable of doing (by way of explaining or
defending his utterance). 'Thinking' has an adverbial character.
For a while, Wittgenstein had continued to identify thought with lan­
guage, albeit 'sign-language' rather than the language of thought: philosophy
is a 'descriptive science . . . of thought'; but thoughts and their logical rela­
tions 'must be examined through the expressions conveying them'; thought
is a 'symbolic process', and thinking 'the activity of operating with signs',
which is performed by the hand, in writing, or the mouth and larynx, in
speaking (LWL 4, 25; BB 6; BT 408). But while we write with our hands, we
only think with them in the sense of accompanying our speech by gestures.
And Wittgenstein himself came to realize that while thinking and speaking
are conceptually related, they are 'categorially distinct' (RPP LI §§6-8, 183-
93, 238, 248, 266-7; Z §§100-3). His mature discussion undermines the
assumption behind both mentalism (Psychokgy II ch. XVIII) and his own
earlier lingualism (of which traces remain in Phiksophkal Investigations §§329-
30), namely that thought requires a medium or vehicle.
His first step was to abandon the catholic use of 'thought', which, like the
mentalist use of 'idea' and 'representation', skates over the differences
between different mental concepts. He treats 'txiinking' as a 'widely ramified
concept', and discusses four major employments (Z §§110-12, 122; RPP LI
§§194, 216): (a) thinking about or meaning something; (b) reflecting on a
problem; (c) believing or opining that p; (d) occurrent thoughts which cross
one's mind at a particular moment. None of them consist in physical or
mental processes, either words or images crossing one's mind, since such
goings-on are neither necessary nor sufficient.
Clearly, long-standing 'convictions' could not consist in images or words
359
THOUGHT/THINKING
constantly crossing one's mind (see PHILOSOPHICAL PSYCHOLOGY). The point
holds also for (b). It would be foolish to deny, as some behaviourists have
done, that when one is flunking mental images may cross one's mind. How­
ever, such inner goings-on are neither sufficient nor necessary for me to
dunk. In a delirium I may have mental images but do not think; and I may
think about a problem without any images crossing my mind. Not all of our
thinking can be characterized as having mental images (a point Berkeley
and Kant made vis-a-vis 'general ideas' or concepts).
The 'linguistic' alternative fares no better. Saying that p and diinking that
p are obviously not the same. Fortunately, we do not express all of our
thoughts in words; and we sometimes say that p when we dunk that q. One
might reply that in such cases we talk to ourselves in foro intemo, and that
thinking is a kind of internal monologue, as Plato had suggested (Theaetetus
189e). But speaking to oneself in the imagination is no more sufficient or
necessary for drinking than is having mental images. When I count sheep in
order to induce sleep, I talk inwardly but I don't think; and one can per­
form even the most complex intellectual tasks without talking to oneself in
the imagination.
This holds even for 'Ughtning-like thoughts' (PI §§318—21). It is implau­
sible to insist that when it suddenly occurs to a motorist: 'You fool; there's a
radar control behind the bridge, you had better slow down to 50!', his mind
runs through that string of words (or of mental images) within a split
second. Mental images and inner speech may be accompaniments of think­
ing, and may be 'logical germs' of thoughts (LW I §843). As psychological
studies in the wake of Vygotsky have shown, they give rise to thoughts and
serve as heuristic or mnemonic devices. However, this dependency is con­
tingent. Inner goings-on do not determine what I think, and are not logi­
cally necessary for me to think. What we think is determined by what we
would sincerely say and do, not by what images or words may flit across
our minds. A motorist can be credited with the aforementioned thought if
he sincerely AVOWS it, either then or later (PI §343; BB 147). Equally, whe­
ther I thought about a problem on a given occasion is determined not by
internal accompaniments, but by what I am capable of doing and by the
way I speak and act, and it may well depend on what went on before or
after.
Wittgenstein also intimates doubt about the idea that when I speak, I
must first dunk in some inner symbolism, linguistic or mental, and then
transpose my thoughts into utterances of a different, public symbolism (BB
41; LPP 247-8). That picture has the absurd consequence that I might
always be mistaken about my own thoughts. For I might read them off
incorrecdy from my internal display of words or images, or mis-translate
them into the sign-language. One can talk inwardly in a particular language,
but this is not the same as to think in a particular language. The question of
360
THOUGHT/THINKING
whether I think in a certain language is simply the question of whether I
need to translate from another language in order to speak this one.
There are essential links between thought and language, but they do not
require any actual inner vocalization. For one thing, we identify thoughts/
beliefs by identifying their linguistic expressions (see BB 4-5, 161; PI §§501-
2; MS108 237). The answer to the question 'What are you diinking?' is not
a description of an inner process, but an expression of my thoughts in
words (e.g., 'I think that it will rain'). If I am challenged by a Platonist or
mentalist to express the thought behind that utterance, I do not re-examine
some inner process to see whether I can describe it better. Instead, I para­
phrase my utterance into other symbols. Consequentiy, language is not just
the only, if distorting, expression of thought, as Frege had it (Posthumous 225,
269-70), it is its ultimate expression. Equally, it is the expression of thoughts
which allows one to speak of their having constituents, as Fregeans do.
The second essential link between thought and language is that the capa­
city for having thoughts or beliefs (c) requires the capacity to manipulate
symbols, not because unexpressed thoughts must be in a language, but
because the expression of thoughts must be. The reason is that ascribing
thoughts makes sense only in cases where we have criteria for identifying
thoughts. Something must count as thinking that p rather than that q. This
means that thoughts, although they need not actually be expressed, must be
capable of being expressed. And only a restricted range of thoughts can be
expressed in non-linguistic behaviour. A dog can think that its master is at
the door, but not that its master will return in a week's time. For it could
not display such a thought in its behaviour (PI §§344, 376-82, 650, II 174;
Z §§518-20). Equally, we can ascribe thinking to, for example, chimpanzees
only because of their problem-solving activities.
James mentions the case of Ballard, a deaf-mute who, having learnt sign-
language, claimed that as a child he had had thoughts like 'What is the
origin of the world?' (Psychology I 266-9). Wittgenstein challenges the idea
that this story provides an empirical proof that thought is possible without
speech. The thrust of his tentative reply (PI §§288, 342; LPP 43) is this. By
contrast to normal cases, whether Ballard thought about the origin of the
world or, for example, about dinner, is not determined by what he could
have said at the time, since ex hypothesi he lacked the ability to use language.
But, Wittgenstein has argued, neither is it straightforwardly determined by
anything which may have crossed his mind. The only possible ground for
attributing a particular thought to him is that he is now translating his pre­
vious wordless thoughts into words. As we have seen, however, in normal
cases there can be no question of having mistranslated one's thoughts, since
there is no such thing as translating one's thoughts into language. But in Bal­
lard's case the question arises whether he has translated his alleged thoughts
correcdy, and this casts doubt on the idea that there was anything to trans-
361
TRACTATUS LOGICO-PHILOSOPHICUS
late to begin with. Ascriptions of thought make sense only on the assump­
tion of expressive abilities, although of course one can be temporarily pre­
vented from exercising them.
Wittgenstein thus links the notion of thinking to potential behaviour
rather than to actual mental goings-on. In some passages, he goes so far as
to question whether thinking is a mental activity (PI §339; RPP LI §193;
MS 124 215). Reflection is not an activity one performs with the brain, since
the latter is not an organ over which one has control. Nevertheless, (a) it is
a voluntary exercise of an acquired mental capacity, just as running is an
exercise of an acquired physical capacity; (b) it can take time, be interrupted
and involve stages; (c) it can be performed in various ways, for example,
with more or less effort; (d) one answer to the question 'What is she doing'
is 'She is thinking about Wittgenstein.' It has been suggested that thinking
per se (rather than thinking about arithmetic, etc.) cannot be taught and does
not consist in anything. But this is equally true of an activity like moving
one's arm. What underlies Wittgenstein's qualms is rather that the different
stages of a thought-process can be identified only by what thoughts the
thinker would vent from one moment to the next, and not by any inner
goings-on. However, this lesson is better expressed by pointing out the differ­
ences between thinking and physical activities (e.g. BB 6-7; RPP LI §217).
Wittgenstein's attack on the language of thought threatens a pillar of con­
temporary cognitive science. It anticipates Ryle's account of the adverbial
nature of tliinking, and his attack on the idea that we must always 'think in'
something (words or images). While avoiding the pitfall of insisting on lan­
guage as the universal medium of thought, Wittgenstein rehabilitates and
radicalizes the Aristotelian idea that HUMAN BEINGS are essentially language-
using animals. Those features that have, at various times, been thought to
distinguish human beings from all other creatures - a capacity for knowing
necessary truths, the possession of a moral sense, self-consciousness or a
sense of history - are all derivative from our distinctive language-using
abilities.
Tractatus Logico-Philosophicus (Logisch-PhUosophische Abhandlung)
Wittgenstein always referred to the only philosophical book he published
during his lifetime as Logisch-Philosophische Abhandlung. Nevertheless, the tide
Moore suggested for die English edition, Tractatus Logico-Philosophicus, has
carried the day and is now an academic household name. Unfortunately,
the work itself has remained obscure. Part of the difficulty, and the appeal
of the book, lies in the fact that it discusses problems like linguistic mean­
ing, the nature of logic, the aim of philosophy and the place of the self, in
a way that combines the formal with the Romantic. "The work is stricdy
philosophical and at the same time literary, but there is no babbling in it'
(FL 10.19). Another obstacle is that the marmoreal remarks of the Tractatus
362
TRACTATUS LOGICO-PHILOSOPHICUS
are extremely condensed. They are not aphorisms, since they are rigidly
fitted into a tight structure. But in his attempt to avoid babbling, Wittgen­
stein adopted a laconic tone and compressed his remarks into what Broad
called 'syncopated pipings'. Wittgenstein himself later acknowledged the jus­
tice of that remark, admitting that every sentence in the Tractatus should be
read as the heading of a chapter, needing further exposition (the back­
ground of the Tractatus's sibylline pronouncements is sometimes provided by
the Notebooks). Finally, the 'decimal numbers assigned to the individual pro­
positions indicate the logical importance of the propositions, the stress laid
on them in my exposition. The propositions n.l, n.2, n.3, etc. are comments
on proposition no. n; the propositions n.ml, n.m2, etc. are comments on
proposition no. n.m; and so on' (TLP In). Wittgenstein considered this
system essential to the book (FL 5.12.19), but it has struck many as mis­
leading. Wittgenstein first used it in the so-called 'Prototractatus', a type­
script which he composed from his Notebooks in 1917-18. Originally, it
served as an aid for composition; but later it turned into a system of sign­
posting. The Tractatus does not apply it consistendy. What Wittgenstein
called his 'basic thought' is tucked away as 4.0312. Propositions 1-7 are
best seen as chapter headings, although 4 is elucidated not by what follows
but by what precedes it.
Wittgenstein had great difficulties finding a publisher for the Tractatus (he
unsuccessfully approached Frege and von Ficker to facilitate publication). It
was eventually published in 1921 in Ostwald's Annalen der Naturphihsophie,
and a year later in an English—German parallel edition. This was thanks to
Russell's generous support. To ensure publication, Russell wrote an Intro­
duction which Wittgenstein condemned as superficial and misleading (RUL
6.5.20), with partial justification.
The work which culminated in the Tractatus started in 1912 as an attempt
to clarify the nature of 'the propositions of logic' and of the LOGICAL CON­
STANTS. Since Wittgenstein explains LOGIC by reference to the nature of
representation, this immediately led on to a 'theory of symbolism' which
elucidates the nature of significant propositions in general (RUL 22.6.12,
26.12.12). The result of the discussion of logic was reached in 'Notes dic­
tated to Moore' (1914), namely that logical propositions are TAUTOLOGIES
which say nothing about reality. Wittgenstein's eventual theory of symbolism
is the PICTURE THEORY (NB 29.10.14), which furnishes the background against
which logical propositions occupy their unique status. Unlike tautologies,
ordinary propositions depict possible states of affairs.
The picture theory draws in its wake an elaborate atomistic ontology of
indestructible OBJECTS. By explaining the essence of the proposition, it
explains the 'essence of being' (NB 22.1.15). This has given rise to a con­
troversy between 'linguistic' interpretations, for which the objects of
the Tractatus are mere posits, and 'ontological! interpretations, for which
363
TRACTATUS LOGICO-PHILOSOPHICUS
language is prior only as regards the ordo cognescendi, not as regards the ordo
essendi. The former is correct in that Wittgenstein's metaphysics is the fallout
of his logic (NL 106): the existence of objects is deduced from a theory of
linguistic representation. However, it is crucial to that theory that language
is not AUTONOMOUS, but must mirror the essential nature of reality in order
to be capable of depicting it. Yet later, Wittgenstein declared that his work
had 'extended from the foundations of logic to the nature of the world' (NB
2.8.16). This not only fits the move from logic to ontology, but also heralds
the emergence of a linguistic version of Schopenhauer's transcendental
idealism (see SOLIPSISM), and of MYSTICAL themes.
The Tractatus comprises four parts, which correspond to stages of its rocky
development: the theory of logic (1912-14), the picture theory (1914), the
discussion of science and mathematics (1915-17), and the discussion of the
mystical (1916-17). The structure of the book is as follows:
Ontology (1-2.063): although the Tractatus is concerned with symbolic
representation (Pref.), it starts with ontology, since the nature of repre­
sentation, and of what represents (thought/language), is isomorphic with
the nature of what is represented (reality).
Depiction (2.1-3.5): having claimed that the world is the totality of facts,
the Tractatus proceeds to investigate a subset of that totality, namely
pictures, in particular PROPOSITIONS, that is, facts which are capable of
representing other facts.
Philosophy (4-4.2): unlike science, philosophy does not consist of proposi­
tions, since the logical form shared by language and reality cannot be
expressed 4n meaningful propositions, but shows itself in empirical pro­
positions (see SAYING/SHOWING).
Theory of logic (4.21-5.641, 6.1-6.13): Wittgenstein uses truth-func-tio-
nal operations to explain the construction of molecular propositions out
of elementary ones - thereby providing an account of the GENERAL PRO-
POSITIONAL FORM - and to establish that logical propositions are tautolo­
gies.
Mathematics (6-6.031, 6.2-6.241): mathematics is also explained as an
aspect of the logical operations by which propositions are derived from
each other.
Science (6.3-6.372): science is treated along Hertzian lines as containing
a priori elements, the network of our description of the world.
Mysticism (6.373-6.522): ETHICAL and AESTHETICAL value is ineffable.
Kicking away the ladder (6.53f.): the Tractatus aims to indicate the limits
of the sayable, but acknowledges that its own pronouncements are on
the far side of the limit. They should be used as a ladder which can be
kicked away once climbed. 'Whereof one cannot speak, thereof one
must be silent' (7).
364
TRUTH
Part of the Tractatus's fascination lies in its elusive unity. A theory of
representation, the picture theory of meaning, delivers an atomistic on
tology; a theory of logic, an account of mysticism and a fascinating picture
of philosophy itself. But some of the links are tenuous. Mysticism is not
ineffable in the same way as logic, and it is not easy to see how the abstract
account of scientific theories fits in with the overall picture of language.
training (Abrichtung) see EXPLANATION
truth There is no theory of truth which has not been ascribed to Wittgen­
stein. He has been 'credited' with a coherence theory, a pragmatic theory, a
consensus theory. The truth of the matter is straightforward. The early Witt­
genstein developed a sophisticated version of the correspondence theory,
while the later Wittgenstein, together with Ramsey, pioneered the redun­
dancy theory. According to the correspondence theory, truth is a relation
between a truth-bearer (judgement, sentence, proposition) and something in
reality which makes it true (a fact). A difficulty for proponents of the theory,
such as Locke, Moore and Russell, is to give a clear account of the notions
of truth-bearer, truth-maker and relation of correspondence. Frege despaired
over specifying a relation of correspondence which would not collapse truth-
bearer and truth-maker. He concluded that truth is sui generis and indefin­
able ('Thought' 59-60).
Wittgenstein tried to meet the challenge. Truth and falsity are not two
abstract entities which the proposition names, as Frege had it (TLP 4.441;
NL 107; see LOGICAL CONSTANTS). Nor are they two properties which proposi­
tions happen to possess, just as roses might happen to be either red or
white, as Russell suggested. Being true and being false are two relations in
which a proposition can stand to reality; and it is an essential feature of
propositions to be BIPOLAR, that is, to be capable of standing in either rela­
tion to reality: a proposition must be capable of being true and capable of
being false. Wittgenstein's positive account starts with observations which
anticipate the redundancy theory and Tarskian theories of meaning and
truth:
(1) '/?' is true = p.
But to understand the proposition that p we need to know more than (1),
namely the logical form of the fact which constitutes the proposition 'p' (NL
104; NM 113).
The PICTURE THEORY provides an account of how a PROPOSITION, which is a
fact, represents other facts truly or falsely. 'A proposition is a picture of rea­
lity: for if I understand a proposition, I know the situation that it represents
365
 TRUTH
. . . A proposition shows how things stand jfit is true. And it says that they do
so stand' (TLP 4.02If.).
Propositions can depict reality either truly or falsely only by being pictures
or models which are compared with reality, as a ruler is laid next to an
object of measurement (TLP 2.1512f., 4.05-4.062; NL 95; NB 24.11.14,
11.1.15). Molecular propositions are truth-functions of ELEMENTARY PROPO­
SITIONS. The truth or falsity of a molecular proposition is determined by the
truth or falsity of its constituent elementary propositions. A molecular pro­
position is true if and only if one of its truth-grounds is fulfilled, that is, if
and only if one of the possible combinations of truth-values under which it
comes out as true in a TRUTH-TABLE actually obtains. Thus, 'p.q' is true if
and only if one of its truth-possibilities obtains, namely the one in which
both 'p' and 'q' are assigned a T in the truth-table.
Elementary propositions are composed of unanalysable NAMES which stand
for simple OBJECTS in reality. Given an appropriate METHOD OF PROJECTION,
the fact that these names are combined in a certain way depicts a state of
affairs, a possible combination or configuration of objects in reality. An ele­
mentary proposition 'p' is true if and only if the state of affairs it depicts
exists, that is, is a fact. This in turn means that the objects for which its
names stand are combined in the way that the combination of nantes in the
proposition says they are. 'A picture agrees with reality or fails to agree; it is
correct or incorrect, true or false . . . The agreement or disagreement of its
sense with reality constitutes its truth or falsity' (TLP 2.21-2.222).
The demise of the ontology of logical atomism in 1929 removed the main
elements of this account. For Austin, this was the stimulus to develop a ver­
sion of the correspondence theory that does not rely on simple objects etc.
For Wittgenstein it was the signal to go back to his starting-point. Having
abandoned the idea that propositions are facts which combine unanalysable
elements, he was left with the simple logical equivalence (1). Ramsey, per­
haps stimulated by Wittgenstein, drew the conclusion that 'It is true that it
is raining' says no more than 'It is raining.' The extra words have no asser­
tive content (Mathematics 138-55). Unlike the early Wittgenstein, he did not
formulate this equivalence through a disquotational statement like (1), but as
(1') 'It is true that p' = 'p'
Wittgenstein followed suit. He insisted that '"p" is true' can be under­
stood only if one treats the sign 'p' as a propositional sign rather than as the
name of a particular ink mark. In contrast to Tarskian theories, Wittgen­
stein rightly denied that 'is true' applies to sentences. Like Ramsey, he had
no qualms about quantifying over propositions, which is necessary to
account for statements that would otherwise defy the redundancy theory.
Thus, 'Whatever the Pope says is true' is rendered as '(p)((the Pope says that
366
TRUTH
p) 3 /»)'; 'What he says is true' is rendered as 'Things are as he says', that is,
'(He says that p).p' (PG 123-4). However, Wittgenstein later reverted to a
disquotational account similar to that adopted by Quine, stating that ' "p" is
true = p' (PI §136; RFM 117). In either version, 'true' does not provide a
peg on which to hang metaphysical disputes, since 'is true' does not state a
relation, either between a proposition and a fact (as realist correspondence
theorists hold), or between a proposition and a set of beliefs (as idealist
coherence theorists have it). But this is not to say that (1) is all there is to
the notion of truth. Wittgenstein discusses at length what it is for different
types of propositions to be true, and what counts as VERIFYING them (OC
§200).
In the course of abandoning logical atomism, Wittgenstein also intimated
a critique of the correspondence theory which prefigures Strawson's later
attack. The correspondence theory treats FACTS as inhabitants of the world.
But facts are not located in space and time, they are neither here nor there.
The fact that the Battle of Hastings was fought in 1066 did not happen in
1066, nor could it have been found on the battle-site. Accordingly, to say
that the proposition that p is made true by the fact that p is misleading, for
there is no extralinguistic item which could do anything to the proposition,
or correspond to it in the way in which a statue and its replica can corre­
spond.
Wittgenstein also denied that one can justify a proposition by pointing to
the fact which, if it obtains, verifies it. One cannot point to (or describe) a
fact, since a fact is not an object or a complex of objects (PR 301-3). All
one can do is to point out a fact. But this is nothing other than to state it.
This means, however, that the verifying fact cannot be invoked as a justifi­
cation, for one would simply be repeating the proposition one was seeking
to justify. One can empirically justify the proposition that p by reference to
the proposition that q. One can also justify it by applying the appropriate
methods of justification successfully. But one cannot justify it simply by
stating that it is a fact that p. "The limit of language is shown by its being
impossible to describe the fact which corresponds to . . . a sentence, without
simply repeating the sentence. (This has to do with the Kantian solution of
the problem of philosophy)' (CV 11; there is indeed a striking parallel with
Kant's 'diallelus' argument, Logik Introd.).
This is not linguistic idealism. Empirical statements are verified or falsified
by the way things are, which is independent of how we say they are. The
truth-value of a proposition is completely independent of grammatical con­
ventions. Yet, to say that propositions are made true by the facts, suggests a
correlation of items like 'Dropping crockery makes it break.' In fact, it is
more akin to saying 'Being a female fox makes one a vixen.' For Wittgen­
stein, it is simply a misleading way of expressing a grammatical proposition,
which is the kernel of truth in the correspondence theory: the proposition
367
TRUTH-TABLES
that p is true if things in fact are as it says they are (BB 30—8; PI §§134,
444).
Wittgenstein does claim that grammar is AUTONOMOUS. But that is a
claim not about truth but about concepts. We must distinguish between
empirical propositions, which are verified or falsified by how things are,
and GRAMMATICAL propositions, which express rules for the use of words.
Rules do not mirror reality, precisely because they cannot be said to be
true or false. Our linguistic practice determines what empirical statements
we can meaningfully make, but not whether these are true or false. Our
conceptual net determines what fish we can catch, but not what, if any,
fish we do catch.
Wittgenstein explicidy denies that a proposition is true if we accept it
(RFM 406; Z §§319, 428-31).
'So you are saying that human agreement decides what is true and what
is false?' - It is what human beings say that is true and false... (PI §241)
Only in thinking do correct and incorrect exist, and hence in the expres­
sion of thoughts: and the expression of thoughts, the language, is
common to men. (MS 124 212-13, quoted by G. P. Baker and P. M. S.
Hacker, Wittgenstein: Rules, Grammar and Necessity (1985), 257)
The words 'is true' have a meaning or role only because human beings
make, dispute and verify assertions; the concept of truth does not exist inde­
pendendy of our linguistic behaviour. But whether or not these assertions
are true depends on how things are, because that is how we use the term
'true'.
truth-tables These are tabular representations of the way in which the
truth-values of molecular propositions depend on the truth-values of their
constituents (ELEMENTARY PROPOSITIONS in the Tractatus). The truth-table is
Wittgenstein's only formal device to have found its way into logic textbooks.
He himself suggests that Frege used truth-tables to explain the logical con­
nectives, but also to make statements about truth-functions. In fact, the idea
goes back to Boole, and the suggestion of using truth-tables as a mechanical
decision procedure was mooted by Peirce and Schroder. It came to techni­
cal maturity simultaneously in Post and the Tractatus (TLP 4.31-4.45, 5.101).
What is unique about the latter is the idea of using truth-tables, not as defi­
nitions of truth-functional connectives, nor exclusively as a decision proce­
dure for the propositional calculus, but as 'propositional signs', a way of
symbolizing molecular propositions, as an alternative to writing them down
as 'p.q' or 'p v q', for example (TLP 4.431, 4.442; AWL 135-6; LFM 177;
cp. Notation §7).
368
TRUTH-TABLES
Truth-tables are a crucial part of the early Wittgenstein's theory of sym­
bolism, his attempt to devise an ideal notation or 'sign-language' (%achen-
sprache) which would reveal the LOGICAL SYNTAX underlying any possible
language. In such a notation, the identity and difference of symbols would
correspond precisely to the identity and difference of things symbolized
(TLP 3.325, 3.342ff., 5.533). Consequentiy, it would show that sentences
which Frege and Russell had treated as different, are one and the same
symbol, alternative ways of writing the same proposition.
Wittgenstein's first attempt to replace what he calls the 'truth-function
notation' of Frege and Russell was the 'annotation' (NL 93-6, 102-3; NM
113-15; LWL 52). A proposition 'p' is written as 'a-p-b', and '~p' as 'b-a-p-
b-a', a and b being the 'two poles' of the proposition, corresponding to T
and F in the Tractatus. What symbolizes in such a formula is the correlation
of the innermost and outermost poles. This shows that 'a-b-a-p-b-a-b' ('—p')
is the same symbol as 'a-p-b' ('/>'), contrary to Frege. Wittgenstein tried to
extend this notation to quantifiers: 'a-(x)-a-<I>*-b-(3x)-b' corresponds to '(x)Ox',
'a-(3x)-a-G>x-b-(x)-b' to '(3x)Ox'. This notation symbolizes internal negation
C(x)~<&x') by reversal of the inner ab poles, external negation ('~(x)Q>x') by
reversal of the outer ab poles. It also shows that the arguments of the quan­
tifiers are propositions with a sense (they have two poles), rather than names
of first-level functions as in Frege. But he encountered insurmountable
obstacles in extending it to IDENTITY (RUL 17.10.13).
In a two-dimensional variation of the annotation, one can display the
connections between the poles of molecular propositions and those of their
constituent atomic propositions (RUL 11.-12.13; NM 115). This provides a
decision procedure for the propositional calculus ('one method'), a mechan­
ical algorithm for distinguishing tautologies, contradictions and contingent
propositions. This cumbersome procedure is included in the Tractatus (TLP
6.1203), but the ai-notation gives way to the truth-table notation (TLP
4.27-4.45, 5.101). A truth-table displays the truth-value, T or F, of a com­
pound proposition for every possible combination of the truth-values of its
constituents (elementary propositions). For a set of n elementary proposi­
tions, there are 2" 'truth-possibilities' or 'truth-combinations', that is, possible
combinations of their truth-values, each represented by a row of the truth-
table. Those truth-possibilities which verify a molecular proposition are its
'truth-grounds'. And there are (2")" 'groups of truth-conditions', one for
every possible truth-function of n propositions. The truth-conditions of a
molecular proposition are its 'agreement and disagreement with the truth-
possibilities of elementary propositions' (TLP 4.431), that is, its truth or fal­
sity for the various truth-possibilities, which is recorded by the last column
of a truth-table. For a pair of propositions p and q there are thus four truth
possibilities, namely T T (both true), FT, TF, FF. Their truth-functions
'p. ~q' and 'p D q\ for example, are represented respectively as:
369
TRUTH-TABLES
' P q
T T F T T T between (non-logical) propositions can be seen from their T / F repre­
FT F FT T sentation (TLP 6.122).
T F T T F F (d)
F F F F F T
'/>. ~q' has a single truth-ground, represented by the third line, (TF), and its
truth-conditions are (FFTF); pZ)q' has three truth-grounds, (TT, (FT),
l
(FF), and its truth-conditions are (TTFT). Unlike their contemporary suc­
cessors, Wittgenstein's truth-tables appear in inverted commas, and without
the proposition at the top of the right-hand column. This indicates that they
neither define the propositional connectives, nor specify the truth-conditions
of molecular propositions, but are themselves propositional signs which
express molecular propositions like '/>. ~q* or 'p Z) q' without recourse to
logical 'constants' or connectives.
Once the order of T's and F's in the first two columns is fixed (it is the
reverse of that in modern textbooks), this notation can be simplified by writ­
ing down the last column as '(FFTF)(/»,o)' or '(TTFT)(/i,ff)' respectively.
Moreover, the elementary proposition '/>' can be represented not just by
'(TF)(/>)', but also by '(TFTF)(/>,o)', that is, as a conjunction of itself with a
tautology involving 'q', for example '/». (q V ~q)'. The T / F column under
this conjunction is identical with the T / F column under 'p' as it occurs in a
2
table with 2 rows (TLP 4.442, 4.465, 5.101, 5.513; NB 3.10.14, 10.6.15).
Accordingly, every proposition can be, and in an ideal notation is, repre­
sented as a truth-function of the entire set of elementary propositions,
namely in a truth-table which, if there are n elementary propositions, has 2"
rows.
For Wittgenstein, the technical innovation of providing a decision proce­
dure was only a means for revealing essential features of logic and symbo­
lism which had been distorted by the formal languages of Frege and Russell.
In particular, the truth-table notation reveals the following essential features
of language:
(a) While genuine propositions have two poles, are BIPOLAR, the proposi­
tions of logic are TAUTOLOGIES which combine bipolar propositions so
that their truth-values cancel each other out.
(b) Molecular propositions are represented through their truth-conditions,
which shows that every proposition is a truth-function of elementary
propositions.
(c) The logical properties of propositions can be calculated (or even
literally seen) from the symbol alone. This replaces the dubious appeal
to the self-evidence of logical propositions by a method for calculating
370
TRUTH-TABLES
the formal properties of symbols. Logical propositions and rules of
LOGICAL INFERENCE become superfluous, since the logical relations
'/>', ' —p' and ' p' are shown to be the same proposition, namely
'(TF)(/>)', which shows that truth-functional connectives do not stand
for functions, but rather express operations (NL 93-4; see LOGICAL
CONSTANTS). In the same manner, '/> V ~p', ~(p.~p)',
c
'p =—p' and
'pZ)p' turn out to be one and the same tautology, '(TT)(/>)'; and
'(/>•(/>Z) q)) D <?' and '{p.(~qZ) ~p)) Z) are similarly both expressed
by '(TTTT)(/>,o)'. This shows that it is impossible to distinguish
between axioms (primitive logical propositions) and theorems.
Although Wittgenstein initially tried to extend the truth-table method to
the predicate calculus, the Tractatus, unlike the Vienna Circle, is perfectly
aware of the fact that the method is restricted to 'cases in which no general­
ity-sign occurs', since it cannot be applied to infinite logical sums or
products (TLP 6.1203; cf. RUL 11.-12./13). For this reason, Church's
theorem, which shows that there can be no decision procedure for the
polyadic predicate calculus, does not direcdy refute the Tractatus. Wittgen­
stein's claim that the logical truths of the predicate calculus are tautologies
in the same sense as those of the propositional calculus was a claim about
the nature of logical truth, not about the scope of a decision procedure.
However, since the idea of a tautology is explained by reference to truth-
tables, the limitations of truth-tables limit the scope of the Tractatus's
account. Two such limitations dawned on Wittgenstein, namely the problem
of explaining GENERALITY in terms of logical products, and the problem of
coLOUR-exclusion. The latter made him realize that not all necessity is tauto­
logical, since there are non-truth-functional logical relations, and he soon
abandoned an attempt to adapt the truth-table notation to such relations
(RLF; WVC 73-4, 91-2). It also forced him to abandon the idea that ele­
mentary propositions are logically independent, which destroys the claim of
the T / F notation to display all meaningful propositions as truth-functions of
elementary propositions, a claim which presupposes that every one of the 2"
rows of the truth-table presents a separate truth-possibility. As a result, in
Wittgenstein's later work, truth-tables lose their paramount role of displaying
the structure of propositions and the nature of logical necessity.
371

u The sense of an elementary proposition is derived from the meanings of its
understanding According to the mentalist theory of meaning epitomized
by Locke, the meaning of a word is an idea, an image in the mind of the
speaker. A similar picture is at work in Russell, for whom understanding a
proposition is being acquainted with what its ultimate consituents stand for,
namely sense-data, and with its logical form. For mentalism, communication
is either a causal process by which speakers induce in their hearers ideas
which are similar to the ones they associate with a word, or a matter of
translation, with speakers translating their internal mental vocabulary into
sounds which their hearers retranslate into their own PRIVATE LANGUAGES.
This position implies that we can never know whether communication has
been successful. Since people are ex hypothesi acquainted with different sense-
data or ideas, they attach 'quite different meanings to their words' (Logic
195; 'Theory' 105-35). So presumably, communication requires only that a
similar idea is produced in the hearer. However, we could never know whe­
ther the speaker manages to do so, since each one of us is acquainted only
with his own ideas.
Frege showed against mentalism that the sense of a sentence, the thought
it expresses, cannot be private, and he concluded that it is an abstract entity
which can be apprehended by different people. However, he was forced to
supplement this Platonist conception of meaning by a mentalist account of
understanding. To understand a sentence is to 'grasp its sense', that is to
latch on to this abstract entity. In communication the speaker does not
induce in the hearer a qualitatively identical idea, but brings him to grasp a
numerically identical thought. Understanding is a 'mental process', albeit
one at the 'very confines of the mental', since it has to cross the ontological
gap between the mental and the abstract. The nature of this process
remains a mystery. It is equally mysterious how we can check whether
speaker and hearer have indeed latched on to the same abstract entity, since
Frege accepts the received idea that the contents of the mind are private
('Sense' 29-30; 'Thought' 68; Posthumous 137-45).
The early Wittgenstein combined Frege's anti-psychologistic evasiveness
with Russell's logical atomism. We are capable of constructing and under­
standing an unlimited number of propositions because we tacitly calculate
372
UNDERSTANDING
their senses from their constituents and their mode of combination. The
sense of a molecular proposition is derived from those of its constituent ele­
mentary propositions according to the rules of truth-functional combination.
unanalysable elements, logically proper NAMES. The process of calculation
presupposes a process of analysis, since the constituents and logical forms of
ordinary propositions are hidden behind their grammatical surface (TLP
3.318, 4.002, 4.024-4.026). Both processes must be unconscious: we are
usually not aware of them, and they will only be made explicit by a success­
fully completed LOGICAL ANALYSIS of the propositions of natural languages.
The result of calculating the sense of a proposition is a string of 'thoughts'
which accompany communication. Thoughts are psychic facts which consist
of thought-constituents that correspond to the names in the propositional
sign. The relation of these constituents to the objects of the depicted situa­
tion 'would be a matter for psychology to find out'. More generally, the
study of 'thought-processes' is irrelevant to logic (TLP 4.1121; RUL 19.8.19;
NB 10.11.14).
Wittgenstein's later approach is radically different. Instead of sweeping the
problem of how we explain and understand words and sentences under the
carpet in the name of anti-psychologism, he develops a non-psychologistic
account of understanding. He rejects the assumption, shared by mentalism
and Platonism, that sentences merely provide the perceptible clothing of lan­
guage-independent THOUGHTS. Frege and the Tractatus were right to regard
mental processes and images as irrelevant to sentence-meaning, but wrong
to think that the notion of meaning can therefore be explained without
reference to the notion of understanding. Communication is not a matter of
making something happen in the hearer's mind, the grasping of a sense,
such that it is irrelevant what happens thereafter. Understanding an utter­
ance is not having an experience, nor is it anything else which happens in
the hearer's mind. Rather, it is an ability, which is manifest in how the
hearer reacts to the utterance (PI §§317, 363, 501-10). Understanding a
word is also an ability, which manifests itself in three ways: in how one uses
the word, in how one responds to its use by others, and in how one
explains what it means when asked (PI §75; AWL 48-50; LFM 19-28).
These three CRITERIA of understanding a word can in principle come apart
(someone might use a word correcdy without reacting appropriately or
being able to explain it), but it is crucial to our concept that they commonly
coincide. Understanding is a 'correlate' of EXPLANATION and meaning, and
instead of asking 'What is the meaning of "X"?' we should ask 'How is "X"
explained?' and 'What are our criteria for someone's understanding "X"?'
(PG 45, 60; BT 11).
During his transition period, Wittgenstein regarded 'understanding' as a
FAMILY-RESEMBLANCE term denoting a variety of interconnected processes (PG
373
UNDERSTANDING
49, 74; PLP 347-8). His rationale was that there are diverse behavioural
manifestations of understanding. Later, this claim recedes into the back­
ground, presumably because he realized that a term may be applied on the
basis of diverse criteria without signifying a family of cases. But he may
have continued to hold that linguistic understanding and other types of
understanding, like understanding people or AESTHETIC understanding, are
connected only by overlapping similarities. For example, one can under­
stand a piece of music without being able to provide a paraphrase. By con­
trast, the understanding of a poem involves a higher degree of linguistic
understanding: one knows how to paraphrase expressions occurring in a
poem, but also why they cannot be replaced by a paraphrase in this context
(PI §§522-35; PG 69; M 105).
Wittgenstein also came to reject the idea that 'understanding' signifies a
family of phenomena (PI §§143-84). Understanding is neither a mental nor a
physical event, process or state. This is not to deny that there may be char­
acteristic mental or physiological 'accompaniments' of understanding, it is to
deny only that these constitute our understanding (PI §152, II 181). Wittgen­
stein adduces three different arguments in favour of this claim. The first is
that no mental or physiological phenomena are logically necessary for under­
standing. Although a variety of images or feelings may cross my mind when
I understand a proposition, none of them are essential to understanding.
Mentalist theories of meaning assume that having a mental image is neces­
sary for connecting an expression and the object it refers to. But this cannot
be a general precondition: otherwise it would be impossible to understand
the order 'Imagine a yellow patch!' without first executing it (PI §§35, 172-
9; BB 12, 149-50). There are physiological prerequisites of understanding -
for example having a brain of a certain size and complexity, or even the
occurrence of specific neural processes. But these are empirical necessities
which tell us nothing about the concept of 'understanding' (BB 7, 118-20;
PI §§149-58, 339, 376; RPP I §1063). Wittgenstein has been accused of
ignoring that neural processes are necessary for understanding, in a meta­
physical rather than an empirical or a conceptual sense. But Wittgenstein
explicitly rejected metaphysical necessities of this kind (see AUTONOMY OF
LANGUAGE). Neither 'Now she understands' (e.g., a word) nor 'Now I can go
on' (e.g., continue an arithmetical series) is a claim about neural goings-on.
The former is based on behavioural criteria, that is, on performance. The
latter is not a description or a report, but an AVOWAL of understanding,
which is not based on evidence of any kind (PI §§151, 179 81, 323). For
others, my sincere avowal is a criterion of my understanding: it usually cre­
ates a presumption that I in fact understand, although that presumption can
be overturned by my failure to manifest this understanding in appropriate
circumstances.
Wittgenstein's second argument is that such phenomena are not sufficient -
374
UNDERSTANDING
their presence does not guarantee understanding. It is tempting to suppose
that having an appropriate mental image guarantees understanding. But if I
am told to fetch a yellow flower, an image of a yellow flower may cross my
mind, without my understanding the order. After all, any mental image
which occurs to me remains to be applied, and there are different METHODS
OF PROJECTING it. Equally, the fact that the correct formula occurs to a pupil
who is taught an arithmetical series does not guarantee that he will be able
to continue the series. This lesson also applies to the Fregean picture. Even
if we grant the mysterious grasping of the sense, how can such an abstract
entity be a 'mode of determining' what the expression stands for? How can
it determine the use of a word over time? It could do so only if it were a
'logical machine', an entity in which all the possible applications were
already laid up in such a way that grasping it takes one through an infinite
number of steps. But this 'philosophical superlative' is sheer mystery-
mongering (PI §§139, 192; PG 40; BB 32-6; LSD 136; see RULE-FOLLOWING).
One might protest that the pupil's understanding consists not in the for­
mula's simply flitting through his mind, but in the fact that he derives his
steps from the formula. Wittgenstein tackles this reply in his discussion of
reading (PI §§156-78). He concedes that the difference between a person
who reads and one who merely pretends to read is that the former derives
what he says from the text. The text is not the cause but the reason for my
reading aloud as I do. Reading is a rule-guided activity. But the attempt to
identify an essence of 'deriving' among the multiplicity of circumstances sur­
rounding it fails. Such failures led James to exclaim that understanding is a
mysterious phenomenon which eludes our coarse psychological vocabulary.
However, this is because we have stripped the onion of its skin in search of
its heart (PI §164; Psychology I 244, 251; this metaphor stems from Ibsen's
Peer Gynt). For whether or not I have derived my words from the text does
itself depend not on anything which went through my mind at the time, but
on what I am capable of doing with the text. Reading is the exercise' of an
ability, not the manifestation of a mechanism, mental or biological.
This conclusion is reinforced by Wittgenstein's third argument. Linguistic
understanding is not an act: it is not something we do, either voluntarily or
involuntarily. Nor is it an event or a process (PI §154; PG 85), since it is not
something which happens or goes on. 'Understanding' signifies an abiding
condition. The moot question is whether it signifies a state, not just in the
sense of being a static verb, but as regards its overall GRAMMAR. Philosophical
Investigations 59n suggests that understanding a word is a state, but not a
mental one, presumably because it is the state of a person rather than a
mind. But other passages repudiate this suggestion (BB 117-18; Z §§71-87)
on grounds Wittgenstein also rehearsed concerning INTENDING AND MEANING
SOMETHING. Unlike mental states (e.g., having a headache), understanding
lacks 'genuine duration': one cannot by checking ascertain whether it is still
375
USE
going on, and it is not interrupted by, for example, sleep. Moreover, there
are no criteria for this state which are independent of its manifestations (PI
§149; see Z §§21, 26, 78, 669; BB 5, 20, 32, 78, 143). This suggests that
understanding is a potentiality rather than an actuality. Moreover, it is not a
disposition, since I do not avow understanding on the basis of observing my
past behaviour under similar circumstances. Rather, linguistic understanding
is an ability (Kbnnen), the mastery of the techniques of using words in count­
less speech activities (PI §150; BT 149; PG 47-51).
The phenomenon of understanding 'in a flash' raises a puzzle for Witt­
genstein's position (PI §§138, 197, H 175-6, 181). The use of a word is
spread out over time and hence it is difficult to see how it could be grasped
in an instant. Wittgenstein replies that the fact that we can understand a
word in an instant is no more mysterious than the fact that in intending to
play chess we do not need to run through all of its rules to ensure that it is
chess and not some other game that we intend to play. 'Now I can go on' is
not the report of an infinitely condensed process (going through the whole
of an arithmetical series), but a reaction. But of course it is not incorrigible.
Whether and what I understand in an instant is determined not by anything
going on at the time, but by what I am capable of doing subsequendy,
which has to conform to an established practice of using the term or con­
tinuing the arithmetical series. And if such reactions of understanding were
generally not followed by successful performance, they would lose their role
in our language-game.
use According to what Wittgenstein called the AUGUSTINIAN PICTURE OF
LANGUAGE, the meaning of an expression is the object to which it refers.
While the early Wittgenstein, along with Russell and Frege, subscribed to a
version of this picture, the later Wittgenstein was the first to subject it to
sustained criticism. Failure to stand for an object does not render an expres­
sion meaningless, and it is a category mistake to treat the object a word
refers to as its meaning. Wittgenstein also presented a famous alternative to
the referential conception. His early work had already given prominence to
the use of signs. However, for the Tractatus the use of a sign merely displays
its combinatorial possibilities, which are determined by the combinatorial possi­
bilities of the object it stands for. It is up to us what NAMES to project onto
what objects, but once we have done so, our use has to mirror the essence
of those referents (TLP 3.326fF., 6.211; NB 11.9.16; see MEANING-BODY).
Wittgenstein's later position is radically different. The meaning of a sign is
not a meaning-body, an entity which determines its use. A sign becomes
meaningful not through being associated with an object, but through having
a rule-governed use (PI §§432, 454; AWL 3, 30). Whether a sign is mean­
ingful depends on whether there is an established use, whether it can be
376
USE
employed to perform meaningful linguistic acts; what meaning it has
depends on how it can be used. 'For a large class of cases - though not for
a l l - i n which we employ the word "meaning" it can be defined thus: the
meaning of a word is its use in the language' (PI §43, cp. §30; BB 69).
Given that Wittgenstein had no qualms about ascribing meaning in this
sense to, for example, proper names, the qualification probably excludes not
certain types of expressions, but a certain sense of 'meaning', namely natural
significance, as in 'These clouds mean rain.'
Wittgenstein's suggestion that meaning is use not only informs the linguis­
tic philosophy of Ryle, Austin and Strawson, but is also accepted by some of
their adversaries (Quine, Dummett), and taken for granted by lexicographers
and field-linguists. It is also plausible: we learn the meaning of words by
learning how to use them, just as we learn how to play chess, not by asso­
ciating the pieces with objects, but by learning how they can be moved.
Nevertheless, it has come in for severe criticism from formal semanticists.
Sometimes, Wittgenstein's followers try to bypass the latter ab initio. Thus, it
is pointed out that he does not proffer a theory of meaning. This is correct,
but does not immunize his position. Wittgenstein elucidates the meaning of
words by describing their use. This presupposes an account of meaning — all
the more so if such investigations are contrasted with systematic theories.
Whether meaning is the sort of thing one should have a theory of depends
on the concept of meaning. On the other hand, Wittgenstein's critics often
ignore that what is at issue here is the ordinary concept of meaning, not
technical notions which formal semanticists might devise.
Another evasive move is to insist that Wittgenstein does not provide even
an account of meaning, but was simply giving a piece of methodological
advice: 'Don't ask for the meaning, ask for the use!' In our investigations of
philosophically contentious terms, the very notion of 'the meaning' misleads
us, since its nominal form suggests an object beyond the sign (this is even
more obvious for Bedeutung, which derives from deuten, i.e. 'pointing'). The
concept of meaning is obsolete save for expressions such as 'means the
same' or 'has no meaning' (M 51-2; AWL 30; PG 56; PI §120). The same
line is taken by Qjiine. However, unlike Quine, Wittgenstein is committed
to the view that philosophical problems about meaning cannot be solved
simply by expunging the term from philosophical vocabulary (see METALOGIC).
Wittgenstein's methodological maxim must be based on a clear under­
standing of the concept of meaning.
Wittgenstein has been blamed for disregarding the fact that the meaning
of a word cannot be identified with particular utterances, or even all actual
utterances, since they include misuses. It has also been objected that we
should be concerned not with how we use words, but with why we use
them the way we do. Both points ignore the normative dimension of Witt­
genstein's conception of linguistic meaning. Wittgenstein elucidates the
377
USE
notion of meaning by establishing its conceptual connections with other
notions, such as UNDERSTANDING and EXPLANATION. The meaning of a word is
what is explained by an explanation of meaning, namely how a word can
be used meaningfully in a particular language. Such explanations are what
Wittgenstein calls GRAMMATICAL rules. They cover an unlimited number of
occasions, and are standards for the correct use of expressions. We invoke
them to justify and criticize our employment of words, which means that
they are our rationales for using words the way we do. And if the question
of why we use words is aimed at establishing the causes of our adopting cer­
tain rules, it is irrelevant to the meaning of the words concerned (although it
may be relevant to their etymology). Meaning is use in accordance with
grammatical rules (AWL 44-8, 85; OC §§61-3).
However, although the notions of meaning and rule-guided use overlap,
they diverge in important respects. First, there are expressions which have a
use, but no meaning, such as 'tally-ho' and 'abracadabra'. Second, unlike its
meaning, the use of a word can be fashionable, accompanied by gestures
and reveal something about the speaker, etc. Third, two expressions may
have the same meaning without having the same use (e.g., 'cop'/'police-
man'). Someone who identifies meaning and use cannot discard these points
as minor details. For they reveal that the use of 'the use of a word' differs
from the use of 'the meaning of a word', and if the identification-slogan
were correct, this would demonstrate that the two words do not mean the
same. The first divergence shows that the notion of use has a wider exten­
sion than that of meaning; the second that there is a category difference
between 'meaning' and 'use'; the third that not all aspects of the use of a
term are relevant to its meaning.
While some passages simply identify meaning and use, others are compa­
tible with the above qualifications (PG 60; LFM 192 vs. PI §139, II 212,
220). Although meaning does not determine use, use determines meaning,
not causally, but logically (just as for Frege sense deteraiines 'meaning', what
a word stands for). While sameness of meaning co-exists with difference of
use, every difference in meaning is a difference in use. Given the use of a
word, we can infer its meaning without further evidence, but not vice versa.
One cannot tell from a dictionary explanation of 'cop' whether the term is
frequendy used by British academics, but one can write the dictionary entry
on the basis of a full description of the term's employment. We can learn
from the use of a word everything there is to its meaning; which means that
conceptual analysis remains a matter of investigating linguistic use. Unfortu­
nately, this does not solve the problem that the term 'use' in vacuo is too
nebulous to be helpful. But it brings the difficulty into focus: what aspects of
use are relevant to meaning? Wittgenstein was aware of this problem. Com­
menting on a fictional language-game in which one and the same type of
tool has a different name on different days of the week, he claims 'not every
378
USE
use is a meaning' (LW I §289). One suggestion, which seems to take us back
to a referential conception, is that what matters about the use of an expres­
sion to its meaning is what it is used of or signifies. However, not all expres­
sions stand for an object. It has been replied that even non-referring
expressions like 'and' signify something: there are 'features' or 'conditions'
that warrant their application. But this amounts to no more than saying that
they are meaningful. The meaning of expressions is a matter of the condi­
tions of their correct use. But that is not disputed by a use-oriented concep­
tion of meaning. We use signs in the world, whether or not we use them to
refer to objects in the world.
Wittgenstein himself suggests that those aspects of the use of a word
which determine its meaning are its role or function, but admits that this
idea is itself imprecise (LW I §§278-304; LPP 291). Elsewhere, he links the
meaning of a word to its purpose or aim, and compares words and proposi­
tions to tools. Important logical differences between words are disguised by
the similarity in their linguistic appearance or form ('2', 'pain', 'table' are all
nouns; 'to swim', 'to mean' verbs) but are revealed by their function, just as
the differences between a hammer and a chisel are revealed in the way they
operate (PI §§5, 11-14, 421, 489; BB 67). However, Wittgenstein did not
hold an instrumentalist conception of meaning according to which the
meaning of a word, like that of a tool, is its effect, namely on the behaviour
of others. Such causal theories had been developed in the twenties by
Russell on the one hand, and Ogden and Richards on the other. For
Russell, speech is a means of producing in our hearers the images which are
in us. The connection between a word and its meaning (an object, or a
mental image of it in the case of memory statements) is a causal one. One
understands" a word actively if suitable circumstances make one use it,
passively if hearing the word causes one to behave in suitable ways (Analysis
ch. X). Ogden and Richards held a similar theory: .the meaning of a symbol
is the thought which hearing it causes or which causes the uttering of it.
Whether the use of a symbol is correct depends on whether it produces in
the hearer a thought similar to that of the speaker.
Both theories were inter alia meant to fill the lacuna left by the Tractatus's
refusal to specify how signs are connected with what they signify. But when
Wittgenstein turned to this problem after his return to philosophy, he criti­
cized Russell, and Ogden and Richards, through arguments which apply to
causal and behaviourist theories in general, and developed his own account
of meaning in direct contrast to them. (For this reason, Quine is wrong to
hold that the idea of meaning as use was anticipated by Dewey, who merely
resisted mentalist theories of meaning in the name of behaviourism.) Just as
causal accounts cannot do justice to the logical nature of INTENTIONALITY,
they cannot do justice to the normative aspect of meaning, and obliterate
the distinction between sense and nonsense. Whether a sign is meaningful or
379
USE USE

meaningless does not depend on whether its utterance has the desired effect, to its linguistic expression, and hence cannot explain the latter's meaning (see
either on a particular occasion or in general. 'This sign means X' does not INTENDING AND MEANING SOMETHING).
mean 'When I utter this sign I get X.' Even if the regular result of my utter­ The conventional or grammatical role of an expression does not just
ing 'Bring me sugar!' is that people stare at me and gape, this does not include the speech acts which can be effected by uttering it (as has been
mean that my utterances mean 'Stare at me and gape!' The meaning of a suggested by those assimilating Wittgenstein to speech-act analysis); it
word is determined by general conventions governing its use, while its effect includes also its combinatorial possibilities, the logical relations of proposi­
depends on contingent conditions pertaining to specific circumstances. tions in which it occurs, and the way in which its employment can be criti­
On the causal theory, there can be no such thing as understanding an cized or justified (see VERIFICATIONISM). Sometimes, Wittgenstein conceives it
order, and yet disobeying it, since in this case the order does not have the too wide, by holding that the meaning of a word is determined by its 'role in
desired effect. A causal theorist might reply that the theory does allow that the whole life of a tribe' (EPB 149). 'Indigestion' has the same role in Eng­
an order can be disobeyed, because it is only one part of the causal chain lish as Kreislaufbeschwerden (circulatory disturbance) in German, namely of
which leads to its execution. In particular, the addressee must be so condi­ expressing the default complaint of hypochondriacs. Yet this does not indi­
tioned as to be willing to obey it. But this reply does not remove the diffi­ cate sameness of meaning, but rather divergence in form of life (paralysis
culty. It is logically possible that the order will be disobeyed even if all the here, Angst there). He is on safer ground in suggesting that the notion of
other links of the chain, including a disposition of the addressee, are func­ sameness or difference in meaning is no more an all or nothing affair than
tioning. A dog, however well trained, may disobey, and a mechanism, how­ sameness or difference in role (PI §§547-70). If this is correct, one should
ever well constructed, may break down. To explain the sense of an order is not try to make the former more precise by invoking the latter, but rest
not to predict its consequences, not even with the proviso that the causal content with distinguishing the conventional role of a word in a language
chain should not be deviant. For it is only by reference to norms of expres­ from its perlocutionary one on a given occasion.
sion (grammatical rules) that we can distinguish between deviant and non-
deviant causal chains, since only the rules determine what counts as comply­
ing with the order or understanding an utterance (PI §§493-8; PR 64; BT
193-4; PG 68-9, 187-92; PLP ch. IV; FW 97).
During the transition period, Wittgenstein claimed that while the meaning
of a word is not identical with its effect, language can be seen as a causal
mechanism lmking stimuli and responses. Later, he rejected this claim, pre­
sumably because it is incompatible with the idea that mechanistic behaviour
which merely happens to accord with a rule does not constitute RULE-
FOLLOWING: if utterances were merely part of a mechanism, they would not
count as moves in a language-game (PI §493; LPP 17, 135, 257). This is not
to deny that language involves causal mechanisms, but rather to say that its
meaningfulness cannot be made intelligible by reference to them.
Even if Wittgenstein did not hold a causal theory, he might have held a
'communication-intention' theory of the kind developed by G. H. Mead and
Grice, according to which the meaning of a word is the effect the speaker
intends to produce by it. But for Wittgenstein what matters to meaning is the
purpose or role of expressions, not speakers (PI §§6, 8, 317, 345). It is not
the intention of speakers to produce a particular effect in their audience by
uttering a form of words in a particular situation (the intention to perform
what Austin calls the perlocutionary act) that matters, but the function an
expression has as a matter of linguistic convention, its role or place in gram­
mar (PG 59, 189-90). Moreover, he would claim that my intention to pro­
duce in hearers a particular effect can itself be understood only by reference

380 381

V conclusively verified or falsified by being compared with reality, because
vagueness see DETERMINACY OF SENSE
variable see PROPOSITION; SAYING/SHOWING; TAUTOLOGY
verificationism This is the view that the meaning of a proposition is its
method of verification (the principle of verification), and that a proposition is
meaningless if it cannot be verified or falsified (the verificationist criterion of
meaningfulness). The principle was first espoused by the Vienna Circle, but
they attributed it to Wittgenstein, who seems to have transmitted it to Wais­
mann in conversations. According to some commentators, the basic contrast
between Wittgenstein's earlier and later work is that between a realist
semantics based on truth-conditions, and an anti-realist semantics which
rejects the notion of verification-transcendent truth and instead setties for
assertability- or justifiability-conditions. The Tractatus indeed states that to
understand a proposition is to know what is the case if it is true (TLP 4.024;
see MEANING). However, this does not mark a simple contrast with verifica­
tionist ideas. Indeed, when Waismann attempted to summarize the Tractatus
in 1930, he moved swiftly from the idea that to understand a proposition is
to know under what conditions one would call it true to the principle of
verification and the verificationist criterion of meaningfulness (WVC 243-5).
This move is at least compatible with the Tractatus: to know whether a pro­
position is true one must verify it, hold it against reality like a ruler (TLP
2.1512f, 2.223). And to understand a proposition is to know what possible
combination of objects would verify it, but not whether that combination
actually obtains. Thus, the Tractatus is tacidy committed to a verificationist
criterion of meaningfulness, although not to the principle of verification.
Wittgenstein started paying attention to the method of comparing a pro­
position with reality in 1929-30, once he realized that a proposition and
what it depicts are not linked through a Iogico-metaphysical isomorphism
(see INTENTIONALITY). As a result, he espoused a full-blown verificationism. A
proposition is meaningful only if it can be verified or falsified completely; its
meaning or sense is the method of verification; a difference of verification is
a difference in meaning; and to understand a proposition is to know how to
382
VERIFICATIONISM
decide its truth or falsity; verification constitutes the whole sense of the pro-
position (WVC 47, 53, 79; PR 66-7, 77, 89, 174, 200; AWL 20; MS 107
143). On this basis, he distinguished between three types of propositions,
according to how they are verified. 'Genuine propositions' (Aussagen) can be
they describe 'primary experience' or sense-data, as in 'It seems as if there is
a sphere in front of me.' They are either true or false. By contrast, 'hypoth­
eses', statements about material objects and the mental states of others, are
not propositions in the same sense, because they are not really true or false,
but only more or less probable. Occasionally, Wittgenstein characterizes
them as laws or rules for constructing genuine propositions (an idea which
he may have got from Weyl): propositions about material objects ('There is
a sphere in front of me') connect multiple propositions about what we see
when looking at them (WVC 100-1, 159, 210-11; PR 282-97). Finally, the
sense of mathematical propositions is given by their proofs. At first, Wittgen­
stein called this a verification of a different kind. Later, he insisted that
proof and experience are not two comparable methods of verification, since
a MATHEMATICAL PROOF does not establish the truth of a statement of fact, but
rather the acceptability of a rule (PR 192, 200; M 60-1; PG 361).
'Genuine propositions' are the successors of the Tractatus's, ELEMENTARY
PROPOSITIONS. They constitute a 'phenomenological language' which is
semantically 'primary'. It is segmented into 'perceptual modality spaces' such
as visual space, auditory space, etc. Hypotheses, that is, everyday proposi­
tions about physical objects and other minds, constitute a 'secondary' lan­
guage, since they are constructed out of phenomenological propositions (see
CRITERIA). The Tractatus had left open the precise nature of elementary pro­
positions, although it suggested that they are about OBJECTS of acquaintance.
Now Wittgenstein adopts an unequivocal phenomenalism. However, he
soon abandoned this position. Initially, a 'phenomenological language' is
semantically primary because it refers to what is immediately given to the
senses (PR 88, 100-4, 267). This is superseded by the idea that what distin­
guishes 'phenomenology' from 'physics' is not reference to something inner,
but that it is purely descriptive, that is, does not provide causal explanations.
Thus understood, 'phenomenology is grammar', the investigation of linguis­
tic rules (BT 437-86; PR 58, 84; WVC 63-8) (although in Remarks on Colour
'phenomenology' refers to a putative subject that claims to stand between
grammar and physics, such as Goethe's theory of COLOURS).
In 1932, Wittgenstein realized that what he had conceived of as the only
genuine propositions are in fact not descriptions of experiences, but AVOWALS.
He also came to reject the idea that hypotheses can never be certain. First,
a proposition can be probable only if it is logically possible for it to be cer­
tain. Secondly, the myth of the given, of private experiences providing the
foundation of knowledge, is undermined by the PRIVATE LANGUAGE ARGUMENT.
383
VERIFICATIONISM VERIFICATIONISM

Thirclly, unlike scientific statements of laws of nature, humdrum propositions dismissed on the popular grounds that we must, rather, distinguish the onto­
about material objects are not rules for the construction of descriptions, but logical question of whether there are verification-transcendent rabbits from
themselves descriptions. the epistemological question of how we could know about them. For Witt­
During the thirties, both Wittgenstein and the Vienna Circle modified the genstein is concerned with the question of whether the purported ontologi­
principle of verification, the latter by conceding that verifying or falsifying a cal statement makes sense. Only if that question can be answered
meaningful proposition need be possible only in principle, and need not be affirmatively, can the question of whether it is true or can be known arise.
conclusive, the former by holding that the method of verification is only one Logic is prior to both ontology and epistemology.
aspect of the sense of a proposition (see USE), and, moreover, that it is one
which does not apply in the case of first-person present tense psychological
propositions. 'Asking whether and how a proposition can be verified is only
a particular way of asking "How d'you mean?" The answer is a contribu­
tion to the grammar of the proposition' (PI §353; BT 265-70; AWL 28-9;
Z §437). Moreover, Wittgenstein gradually realized that not all aspects of
the method of its verification are part of the sense of a proposition, but only
those which are linked to the way the relevant concepts are explained. In
1932-3, he argued that the fact that we can learn about who won the boat
race by reading a newspaper goes some way to explaining the meaning of
'boat race'. Later, he insisted that to say that the length of playing fields is
measured with the help of tripods is a matter of physics, while to say that
measuring involves the possibility of comparing the lengths of different
objects is partly constitutive of the meaning of 'length' (M 59-60; PI II 225).
Unlike contemporary anti-realism, Wittgenstein never cast doubt on the
intelligibility of empirical propositions which are verification-transcendent
but for which there can be evidence of some sort, such as propositions
about the past (e.g., concerning Rosa Luxemburg's last thoughts) or the
future ('A city will never be built here'). But he questioned the intelligibility
of metaphysical propositions which are such that nothing would count as
evidence for or against them. This holds, for example, for the sentence
'There is a white rabbit between the chairs whenever no one observes
them', but equally for Russell's suggestion that 'The world might have been
created five minutes ago, complete with records of the past' (LWL 111;
AWL 25-6; BB 45-6; PI II 221). However, this is not a lazy philistinism
which accepts only problems for which we have answers. Wittgenstein's
point is not epistemological, namely that we could never know, but logical,
namely that such propositions are 'idle wheels'. These strings of words
cannot be used to make a move in the language-game, if they are taken in
the way intended by the metaphysician. Metaphysical uses of words like
'flux', 'vagueness' or 'appearance' are without 'antithesis'. The metaphysi­
cian is not prepared to count anything as stable, accurate or real. That
means, however, that he has not explained what his contrast between
apparent and real amounts to. There are no standards of correctness for his
metaphysical use of these terms, and hence his employment of them is
meaningless. Whether or not this verdict can be sustained, it cannot be

384 385

w
will Wittgenstein's early treatment is influenced by Schopenhauer's idea
that the world as it appears to us is a manifestation of an underlying reality,
an impersonal, cosmic will. We can know this noumenal reality, since our
bodies are direct manifestations of it (not mere phenomena), and since we
have access to our own willing, the only event we understand 'from within',
not merely as a phenomenon (World I §19, II ch. 18).
Wittgenstein's discussion of SOLIPSISM is based on a Schopenhauerian dis­
tinction between the illusory 'thinking subject', and a metaphysical self
which is not only the ineffable subject of experience, but also the 'willing
subject' (NB 2.-5.8.16; TLP 5.631). He also distinguishes between 'the will
as a phenomenon . . . of interest only to psychology' and 'the will as the sub­
ject of ethical attributes'. The former is part of the episodes which constitute
an individual's mental life, the latter is housed in the metaphysical self, and
hence ineffable (TLP 6.423; NB 21.7.16). Like Schopenhauer, Wittgenstein
views the world as morally inert and locates ETHICS in this metaphysical will.
But for Schopenhauer salvation lies in overcoming the dictates of this blind
force, while Wittgenstein regards the will as the 'bearer' of both 'good and
evil' (NB 21./24./29.7.16). As in Schopenhauer, the metaphysical will is
impersonal and 'permeates' the world, although this 'world-will' is 'in a
higher sense my will' (NB 11.6./17.10.16). At the same time, Wittgenstein
rejects Schopenhauer's metaphysics of the will as a thing in itself of which
the phenomenal world is a manifestation. The metaphysical will is not a pri­
mordial force operating in the world, but an ethical 'attitude of the subject
to the world'. It does not alter the facts but rather 'the limits of the world'
(NB 5.7./4.11.16), namely the transcendental self's attitude towards the facts
which constitute the world - an idea reminiscent of Kierkegaard.
Underlying this position is the view that 'the world is independent of my
will', that I am 'completely powerless' to bend events to my will (TLP 6.373;
NB 11.6-/8.7.16). One possible reason is that my only relation to the world
which matters to logic is my depicting it in propositions. Yet, according to
Tractatus 6.423, there is a psychological difference between different proposi­
tional attitudes like thinking and willing that p. As an empirical phenom-
386
WILL
enon, however, the will is impotent in a sense which is crucial to Wittgen­
stein's early position. There is no logical relation between the occurrence of
any two empirical events, but only a contingent one (see CAUSATION), and this
holds equally for my willing that p and p's coming about. This has three
important consequences. First, the freedom of the will consists merely in the
fact that we cannot know, that is, logically infer, our own future actions
(TLP 5.135-5.1362). Second, if what we 'wish' happens, this is only a con­
tingent 'physical connection', which itself is not under my control (TLP
6.374). By the same token, although there is a difference between those
parts of my body which are and those which are not under my control
(TLP 5.631), that control is merely a contingent one. This means, finally,
that, pace Schopenhauer, I do not have an intuitive certainty of my inten­
tional actions; my body is a mere phenomenon, on a par with all other
parts of the world (TLP 5.641; NB 2.9./12.10./4.11.16).
Thus, the Tractatus presents a contemplative conception of the will: the
phenomenal will is an ordinary empirical event, which merely happens to
us, and is only contingentiy related to our actions; the transcendental will is
a mere ethical perspective. Certain passages in the Notebooks put pressure on
this paradoxical position. Firsdy, thinking itself involves an exercise of the
will, and may be impossible without our controlling at least certain mental
events (NB 21.7.16). This intimates a major difficulty in Wittgenstein's posi­
tion, which insists on the impotence of the will while relying on the trans­
cendental will for connecting language with reality through something akin
to mental ostension. Secondly, there is a difference between wishing and
willing. The former is indeed merely a mental phenomenon that may or
may not be followed by a bodily movement. The latter, however, is not
contingently related to action, it 'is acting', the volition is 'the action itself.
Hence it can involve certainty (I can predict that T shall raise my arm in
five minutes) and a feeling of responsibility. By the same token, the relation­
ship between volition and act is not that of cause and effect. This is pre­
cisely Schopenhauer's position, as is the claim that 'the act of will is not an
experience' (NB 4.-9.11.16; see World I §18).
When Wittgenstein later developed this point it was first through the idea
that representation itself involves IOTENTIONALITY. Like diinking, willing is not
a phenomenon that 'simply happens' and which we observe 'from outside',
but something 'we do'; it consists in our being 'in the action', as its true
'agent' (PG 143—50). Philosophical Investigations moves on to discuss in its own
right the concept of willing, perhaps because of the importance it attaches to
human practice, but also because of conflicting philosophical pressures.
Wittgenstein's aim is to undermine both the empiricist idea that 'willing too
is merely an experience', and the 'transcendental' idea of willing as an
'extensionless point', an ineffable mental force (PI §§611, 620; EPB 236).
The conflict has two interrelated dimensions, the question of whether willing
387
WILL
is something beyond our control, and the question of whether willing is a
mental accompaniment of action, as wishing is.
The empiricist position attacked is that of the Tractatus, but also of Russell
(Analysis ch. XTV), and especially James (Psychology I), whose ideo-motor
theory conceives of willing an act as the occurrence of prior kinaesthetic
sensations, and assimilates willing to wishing. In both James and the early
Wittgenstein the idea that willing is a phenomenon over which we lack con­
trol is fuelled by experiments like the intertwining of fingers or drawing from
a mirror image. They suggest that one cannot produce the requisite experi­
ence of willing prior to the action. It seems that willing 'comes when it
comes; I cannot bring it about', that 'one can't will whenever one wants. It
just happens' (EPB 235-6; cf. NB 4.11.16; BB 153-5; PI §§612, 617). But if
I am unable to bring about my willing, I would be powerless even if the
connection between volition and action were one of necessity.
Against the view that willing is an experience which I cannot bring about,
Wittgenstein makes the following points: (a) it is only in special cases, for
example in the absence of muscular effort, that we say of an action that 'it
comes when it comes'; (b) in the ordinary sense of 'bring about' it is possible
for me to bring about, say, my willing to swim, namely by jumping into the
water - we learn how to will to O by learning to O; (c) the fihger-mtertwin-
ing and mirror-drawing experiments contrast with the case where the finger
is prevented from moving merely by being held, they involve not an inabil­
ity to will but an inability to find any point of application for the will; (d)
the relation between willing and bodily movement is not merely contingent,
as the Tractatus had it: 'when "I raise my arm", my arm goes up' (PI §§612-
21; EPB 236).
However, the empiricist position need not stress the idea of impotence. It
may be looking in experience precisely for the real doing, the real agent.
This search lies behind the famous question 'what is left over if I subtract
the fact that my arm goes up from the fact that I raise my arm?' (PI §621).
That the exercise of the will is phenomenally distinguishable in experience is
suggested by the fact that I can be certain as to whether I am willing, and if
so, what I am willing (NB 4.11.16): how could I know this unless the willing
and its content could be read off from my experience? Wittgenstein dis­
cusses two candidates for the experience constituting the real volition or
doing (PI §§621—6; BB 51). James suggested that a voluntary physical action
feels different from an involuntary movement, since it involves KINAESTHETIC
sensations of action. Wittgenstein retorts that much of our authority to pro­
nounce about ourselves does not involve kinaesthetic feelings, and even
when it does they do not underpin judgements of having acted voluntarily,
since we identify those feelings by reference to the voluntary movements of
our limbs.
The second candidate for a phenomenal doing has been popular in
388
WILL
recent action theory, namely, trying. Wittgenstein rejects this candidate on
the grounds that not all actions involve trying. He suggests that it is false to
say that I try to O if my Oing involves neither an effort nor a possibility of
failure. To this Griceans have objected that although we are reluctant to
speak of 'trying' in such cases, the reason is not that this would be false, but
that it would be too obvious to be worth stating. But their position does not
accord with the linguistic facts. It is committed to the mystifying claim that
it is less obvious (and hence more worth saying) that I am trying to <I> when
my Oing involves an effort. Moreover, if an interlocutor were to tell one
'Graf is trying to play tennis' in a situation in which Graf is effortlessly
applying her forehand, one would not respond 'No need to tell me, I can
see that she is.' Rather, one would react to the statement as a misapplica­
tion of a word: 'What do you mean "She's trying", can't you see how
effortlessly she is playing?'
For Wittgenstein, the failure of these attempts to identify a phenomenon
of willing is no coincidence. As soon as we try to identify the real doing
with something in experience, it will appear as a mere phenomenon, some­
thing itself produced, not the unmoved mover behind the action. 'The will
can't be a phenomenon, for whatever phenomenon you take is something
that simply happens, something we undergo, not something we do' (PG 144).
But the idea that the empiricist picture makes a nonsense of agency is the
sole motive behind this transcendental alternative, which locates the real
agent beyond experience.
Wittgenstein rejects this picture as equally awry. There are experiences
involved in voluntary action (e.g., our seeing and feeling that we raise our
arm). When we try to 'distinguish between all the experiences of acting plus the
doing (which is not an experience) and all those experiences without the ele­
ment of doing', the element of doing appears 'redundant' (PG 145). Nothing
is left over in experience when we subtract the experience of our arm rising
from the experience of our raising our arm. But this does not show that
there is a real doing left over which is not in experience. Willing, unlike
wishing, is not a mental event prior to or accompanying the bodily action.
It is the action, as the Notebooks suggested, not, however, in a mysterious
Schopenhauerian sense, but in 'the ordinary sense' of speaking, writing,
walking, etc. And in cases in which I try but fail to O, it is the trying to O
(PI §§614-16).
This denial that willing is a mental accompaniment of action parallels
Wittgenstein's account of thinking (see THOUGHT/THINKING). The difference
between voluntary and non-voluntary movement does not lie in mental
goings-on, but in the context, and in what the agent is capable of doing on
this occasion. Wittgenstein mentions the following features of voluntary
action (Z §§577-99; PI §§611-28; BB 157): (a) susceptibility to orders, and
the manner of this susceptibility - orders are normally not obeyed auto-
389
WILL
matically; (b) the possibihty of deciding whether or not to <X>; (c) the char­
acter of the movements and their relations to other surrounding events and
circumstances; (d) different conclusions we draw from them, notably con­
cerning responsibility; (e) while one can wish for anything, one can will only
what is, or what one believes to be, within one's powers.
If the distinction between the voluntary and the involuntary does not he
in the presence or absence of an extra element of willing, willing is not the
source of our voluntary actions. This undermines the idea that willing is our
executive relation to our physical acts, an idea shared by both empiricism
and transcendentalism. There are two important consequences. First, we do
not use any means to bring about our action, for example an act of wishing
(PI §614). Second, the conflict between empiricism and transcendentalism is
based on a wrong assumption. Willing is neither a caused event which hap­
pens to me, beyond my control, nor 'an immediate, non-causal bringing-
about' (PI §613). Wittgenstein reinforces this conclusion by arguing that will­
ing is neither voluntary nor involuntary: (a) it makes no sense to speak of
'willing willing'; (b) if it did, 'willing' would be the name of an act, the act of
willing, but it is not — one cannot, for example, obey the command to will,
as nothing is specified thereby; (c) it makes sense to say that my body does
not obey my will, but not that my will does not.
This line of thought is reminiscent of Ryle's argument against volitions
and causal conceptions of the will. The insight that we don't usually cause
our own actions has echoes in Davidson. But the claim that the empiricist
position makes a nonsense of voluntary action by turning the will into a
mere phenomenon, and the denial that particular actions are inherently
voluntary due to a special origin, run counter to the causal accounts of the
mind forcefully developed by Davidson. Regarding the problem of free will,
Wittgenstein, like Schopenhauer, denied that libertarianism is vindicated by
an experience of free volitions. He tried to avoid determinism by claiming
that the fact that our actions follow natural laws does not show that we are
in any way 'compelled', but his cursory reflections on this matter (LFW) are
themselves uncompelling.
390
Bibliography of secondary sources
This survey of the secondary literature is extremely selective. For further
bibliographical assistance see
Frongia, G. and McGuinness, B., Wittgenstein: A Bibliographical Guide (Oxford:
Blackwell, 1990).
Shanker, V. A. and Shanker, S. G., A Wittgenstein Bibliography (Beckenham:
Croom Helm, 1986).
[The latter is more comprehensive, the former features useful abstracts.]
HISTORICAL BACKGROUND
Hacker, P. M. S., Wittgenstein's Place in Twentieth-Century Analytical Philosophy
(Oxford: Blackwell, formcoming). [A magisterial if partisan treatment.]
Haller, R., Questions on Wittgenstein (London: Roudedge, 1988). [Sheds valu­
able light on the Germanophone context of Wittgenstein's work.]
Janik, A. and Toulmin, S., Wittgenstein's Vienna (New York: Simon & Schus­
ter, 1973). [Illuminates Wittgenstein's intellectual milieu, although many
of its exegetical claims are problematic]
Monk, Ray, Wittgenstein: the Duty of Genius (London: Cape, 1990).
McGuinness, B., Wittgenstein, a Life: Young Ludwig 1889-1921 (London: Pen­
guin, 1988). [Both biographies are excellent, McGuinness is stronger on
Wittgenstein's intellectual background, Monk on his character.]
Nedo, M. and Ranchetti, M., Wittgenstein: San Leben in Bildem und Texten
(Frankfurt: Suhrkamp, 1983). [A lavish and well-documented collection of
photographs].
Passmore, J., A Hundred Years of Philosophy (London: Duckworth, 1966). [The
most comprehensive history of analytic philosophy - a masterpiece.]
Skorupski, J., English-Speaking Philosophy 1750-1945 (Oxford: OUP, 1992).
[The tide notwithstanding, the book deals with the development of philo­
sophical modernism in Frege, Russell and Wittgenstein.]
Urmson, J. O., Phiksophkal Analysis (Oxford: OUP, 1956). [A bit dated on
TLP, but strong on Cambridge between the wars.]'
Wedberg, A., A History of Phiksophy, Vol. 3: From Bolzano to Wittgenstein (Ox­
ford: OUP, 1984).
391
 BIBLIOGRAPHY OF SECONDARY SOURCES
GENERAL INTRODUCTIONS
Fogelin, R. F., Wittgenstein (London: Roudedge, 1987; 1st edn. 1976). [Ex­
cellent on TLP; contains the first statement of the rule-sceptical inter­
pretation of PI. The diagram on p. 109 and the list on p. 153 are derived
from this book.]
Grayling, A. C., Wittgenstein (Oxford: Oxford University Press, 1988). [A
brief introduction to the main themes.]
Hacker, P. M. S., Insight and Illusion (Oxford: Clarendon Press, 1986; 1st
edn. 1972). [Arguably the best single book on Wittgenstein; traces the de­
velopment of his views on philosophy and the mind.]
Kenny, A., Wittgenstein (Harmondsworth: Penguin, 1973). [Still the best first
introduction; excellent on TLP; stresses the continuity between the early
and later work.]
Malcolm, N., Nothing is Hidden: Wittgenstein's Criticism of his Early Thought
(Blackwell, Oxford, 1986). [An excellent account of the contrast between
the early and later views.]
Pears, D., Wittgenstein (London: Fontana, 1971). [Places Wittgenstein in the
tradition of critical philosophy.]
Schulte, J., Wittgenstein: an Introduction (New York: State University of New
York Press, 1992; German edn. Stuttgart: Reclam, 1989). [Covers a wide
range of topics in a very accessible manner.]
COMMENTARIES
Black, M., A Companion to Wittgenstein's Tractatus' (Cambridge: Cambridge
University Press, 1964). [Provides a wealth of background material; less
good at shedding light on difficult passages.]
Baker, G. P. and Hacker, P. M. S., Wittgenstein: Understanding and Meaning -
Vol. 1 of An Analytical Commentary on the Philosophical Investigations (Oxford:
Blackwell, 1980).
Wittgenstein: Rules, Grammar and Necessity - Vol. 2 of An Analytical Com­
mentary on the Philosophical Investigations (Oxford: Blackwell, 1985).
Hacker, P. M. S., Wittgenstein: Meaning and Mind - Vol. 3 of An Analytical Com­
mentary on the Philosophical Investigations (Oxford: Blackwell, 1990).
Wittgenstein: Mind and Will - Vol. 4 of An Analytical Commentary on the
Philosophical Investigations (Oxford: Blackwell, 1996). [The best commen­
taries on PI; combines scholarship with rigorous argument.]
Hallett, G., A Companion to Wittgenstein's 'Philosophical Investigations' (Ithaca: Cor­
nell University Press, 1977). [Unlike the other commentaries, Hallett also
covers PI Part II.]
Savigny, E. von, Wittgensteins Philosophische Untersuchungen: Ein Kommentar fur Le-
ser (Frankfurt: Klosterman, 1988). [Interprets PI without recourse to the
Nachlass; close scrutiny of linguistic details.]
392
BIBLIOGRAPHY OF SECONDARY SOURCES
COLLECTIONS OF ESSAYS
Arrington, R. L. and Glock, H.-J. (eds), Wittgenstein's Philosophical Investigations:
Text and Context (London: Roudedge, 1991). [Essays devoted to particularly
difficult passages in PL]
(eds), Wittgenstein and Quine (London: Roudedge, forthcoming). [Discus­
sions of the similarities and differences between the two.]
Block, N. (ed.), Perspectives on the Phiksophy of Wittgenstein (Oxford: Blackwell,
1981). [An excellent collection.]
Canfield, J. (ed.), The Phiksophy of Wittgenstein: A Fifteen Volume Collection (New
York: Garland, 1986). [The most comprehensive collection of critical es­
says.]
Copi, I. M. and Beard, R. W. (eds), Essays on Wittgenstein's Tractatus (London:
Roudedge, 1966). [Collects the best of the early essays on TLP.]
Fann, K. T. (ed.), Ludwig Wittgenstein, the Man and his Phiksophy (Hassocks:
Harvester, 1967). [Contains assessments by contemporaries.]
Glock, H.-J. (ed.), Wittgenstein - A Critical Reader (Oxford: Blackwell, forth­
coming). [These essays are intended to accompany the selections in A.
Kenny, The Wittgenstein Reader]
Griffiths, A. P. (ed.), Wittgenstein: Centenary Essays (Cambridge: CUP, 1991).
[Contains some essays on unusual topics.]
Heringer, H. J. and Nedo, M. (eds) Wittgenstein and his Times (Oxford: Black-
well, 1982; German edn. 1979). [Essays on Wittgenstein's perspective on
modernity.]
Hintikka, J. (ed.), Essays on Wittgenstein in Honour of G. H. von Wright (Am­
sterdam: Norm-Holland, 1976). [Contains important essays, mcluding
Cioffi on aesthetic explanation.]
Kenny, A., The Legacy of Wittgenstein (Oxford: Blackwell, 1984). [Incisive and
profound essays on Wittgenstein's work and its impact.]
Klemke, E. D. (ed.), Essays on Wittgenstein (Urbana: University of Illinois
Press, 197,1).
Luckhardt, G. (ed.), Wittgenstein: Sources and Perspectives (Ithaca: Cornell Uni­
versity Press, 1979). [A valuable source for scholars.]
Malcolm, N., Wittgensteinian Themes (Cornell: Cornell University Press, 1995).
Pitcher, G. (ed.), The Phiksophkal Investigations (London: Macmillan, 1968).
[Seminal essays from the early period of Wittgenstein interpretation.]
Schulte, J., Chor und Gesetz (Frankfurt: Suhrkamp, 1990). [Highly informative
essays on topics which are usually neglected.]
Shanker, S. (ed.), Wittgenstein: Critical Assessments, 4 vols (London: Croom
Helm, 1986). [Contains most important exegetical essays.]
Teghrarian, S. (ed.), Wittgenstein and Contemporary Phiksophy (Bristol:
Thoemmes, 1994). [Most of the essays discuss Wittgenstein's relationship
with contemporary philosophy.]
393
BIBLIOGRAPHY OF SECONDARY SOURCES
Vesey, G. (ed.), Understanding Wittgenstein (Ithaca: Cornell University Press,
1974). [An important collection of essays.]
Winch, P. (ed.), Studies in the Philosophy of Wittgenstein (London: Roudedge &
Kegan Paul, 1969). [Cook's 'Human Beings' is particularly valuable.]
von Wright, G. H., Wittgenstein (Oxford: Blackwell, 1982). [An invaluable
collection, features essays on the origin of TLP and PI, as well as a cata­
logue of the Nachlass]
THE EARLY WORK
Anscombe, G. E. M., An Introduction to Wittgenstein's Tractatus (London: Hutch­
inson, 1959). [Difficult, but vigorous and stimulating.]
Baker, G. P., Wittgenstein, Frege and the Vienna Circle (Oxford: Blackwell, 1988).
[A thorough and subde treatment of the development of Wittgenstein's
account of logical necessity.]
Carruthers, P. Tractarian Semantics (Oxford: Blackwell, 1989).
The Metaphysics of the Tractatus (Cambridge: CUP, 1990).
[Both books assimilate TLP to Frege in a problematic fashion, but feature
interesting criticisms of orthodox interpretations.]
Griffin, J. P., Wittgenstein's Logical Atomism (Oxford: Clarendon Press, 1964).
Lange, E. M., Wittgenstein und Schopenhauer (Cuxhaven: Junghans, 1989).
[Sheds light on the background of TLP's treatment of solipsism.]
McDonough, R. M., The Argument of the Tractatus (Albany: State University of
New York Press, 1986).
Mounce, H. O., Wittgenstein's Tractatus: an Introduction (Oxford: Blackwell,
1981). [The most accessible introduction to TLP.]
Pears, D., The False Prison, Vol. / (Oxford: Clarendon Press, 1987). [Features
an illuminating account of the emergence of the picture theory from Rus­
sell's theory of judgement.]
Stenius, E., Wittgenstein's Tractatus (Oxford: Blackwell, 1960). [The first realist
interpretation of TLP; stresses Kantian aspects.]
THE LATER WORK
Baker, G. P. and Hacker, P. M. S., Scepticism, Rules and Language (Oxford:
Blackwell, 1984). [A vehement attack on the rule-sceptical interpretation.]
Barrett, C , Wittgenstein m Ethics and Religious Belief (Oxford: Blackwell, 1991).
[A survey of Wittgenstein's views.]
Budd, M., Wittgenstein's Philosophy of Psychology (London: Roudedge, 1989).
[Meticulous and level-headed account.]
Canfield, J., Wittgenstein: Language and World (Amhurst: University of Massa­
chusetts Press, 1981. [A painstaking discussion of the notions of grammar
and of criteria.]
394
BIBLIOGRAPHY OF SECONDARY SOURCES
Diamond, C., The Realistic Spirit (Cambridge, Mass: MTT, Press, 1991). [Dif­
ficult but interesting, especially on the topic of nonsense.]
Fann, K. T., Wittgenstein's Conception of Philosophy (Oxford: BlackweE, 1969).
Frascolla, P., Wittgenstein's Philosophy of Mathematics (London: Roudedge, 1994).
[Identifies unifying themes in Wittgenstein's early and later philosophy of
mathematics.]
Harming, O., Wittgenstein's Later Philosophy (London: Macmillan, 1989). [Lucid
and level-headed, focuses on the idea that explanations come to an end;
the diagram on p. 121 is taken from Hanfling.]
Hark, M. ter, Beyond the Inner and the Outer (Dordrecht: Kluwer, 1990).
Hilmy, S., The Later Wittgenstein (Oxford: Blackwell, 1987). [An immensely
scholarly investigation of the emergence of Wittgenstein's later views.]
Hintikka, M. B. and Hintikka, J., Investigating Wittgenstein (Oxford: Blackwell,
1986). [Interprets the private language argument as the result of Wittgen­
stein's abandonment of phenomenalism.]
Hunter, J., Understanding Wittgenstein (Edinburgh: Edinburgh University
Press, 1985). [Short but highly iUurninating essays on specific passages of
PI]
Johnston, P., Wittgenstein and Moral Philosophy (London: Roudedge, 1989). [A
sympathetic account of Wittgenstein's views on ethics and moral psychol­
ogy-]
Wittgenstein: Rethinking the Inner (London: Roudedge, 1993).
Kripke, S. A., Wittgenstein on Rules and Private Language (Oxford: Blackwell,
1982). [A powerful presentation of rule-scepticism and the community-
view of rule-following.]
Marion, M.,-Wittgenstein and Finitism (Oxford: Clarendon Press, forthcoming).
McGinn, C , Wittgenstein on Meaning (Oxford: Blackwell, 1984). [A good dis­
cussion of rule-following and understanding.]
Mulhall, S., On Being in the World (London: Roudedge, 1990). [An iUurninat­
ing comparison of Heidegger and the later Wittgenstein.]
Pears, D., The False Prison, Vol. II (Oxford: Clarendon Press, 1988). [Focuses
on the private language argument and rule-following; difficult.]
Phillips, D. Z., Wittgenstein and Religion (Basingstoke: St Martin's, 1993).
[Essays by the most eminent proponent of a Wittgensteinian philosophy
of religon.]
Rundle, B., Wittgenstein and Contemporary Philosophy of Language (Oxford: Black-
well, 1990). [Incisive and subde criticisms of Wittgenstein's later philoso­
phy of language.]
Schulte, J., Experience and Expression (Oxford: Clarendon Press, 1993; German
edn. 1987). [Discusses Wittgenstein's philosophical psychology after PI;
the diagrams on pp. 290-1 are developed from Schulte.]
Shanker, S., Wittgenstein and the Turning Point in the Philosophy of Mathematics
395
BIBLIOGRAPHY OF SECONDARY SOURCES

(London: Croom Helm, 1987). [The best defence of Wittgenstein's philo­

sophy of mathematics.]
Stroll, A., Moore and Wittgenstein on Certainty (Oxford: Oxford University Press,
1994). [Stresses that for OC the basis of certainty lies in actions rather
than propositions.]
Wright, C , Wittgenstein on the Foundations of Mathematics (London: Duckworth, Index
1980). [The first book to examine the implications of the rule-following
considerations for the account of logical necessity.]

WITTGENSTEIN'S IMPACT ON OTHER DISCIPLINES

Entry tides are printed in small capitals. Bold page-numbers indicate a sus­
Baker, G. P. and Hacker, P. M. S., Language, Sense and Nonsense (Oxford: tained discussion of a topic, whether or not it features as an entry.
Blackwell, 1984). [Uses Wittgenstein's ideas to criticize theories of mean­
AESTHETICS 18, 3 1 - 5 , 123, 251-3; and Austin, J.L. 77, 80, 128, 366, 377, 380
ing in contemporary philosophy and linguistics.] ETHICS 28, 3 1 - 2 , 107 AUTONOMY OF LANGUAGE/ARBITRARINESS
Bloor, D., Wittgenstein: A Social Theory of Knowledge (New York: Columbia Uni­ ability see disposition OF GRAMMAR 2 2 , 4 5 - 5 0 , 7 4 , 8 1 , 8 4 ,
versity Press, 1983). [Exploits Wittgenstein for the sociology of knowl­ abstractionism 314 174, 202, 239-41, 275-7, 295-6, 336,
edge.] acquaintance 43, 102, 160, 208, 212, 352, 368
254, 269-70, 277, 299, 310, 348-9, AVOWAL (Aujlerung/Ausdruck) 23, 27, 45,
Bouveresse, J., Wittgenstein Reads Freud (Princeton: Princeton University Press,
383 5 0 - 4 , 56, 88, 144, 162, 175-6, 181-4,
1995). [An evenhanded account of the implications of Wittgenstein's re­ agreement 128-9, 135-6, 328, 368 309, 374
marks for our understanding of the unconscious.] ambiguity/synonymy 40, 122, 240
Chapman, M. and Dixon, R. A. (eds.), Meaning and the Growth of Under­ analytic/synthetic 18, 20, 131, 199-200, BaUard-case 3 6 1 - 2
standing: Wittgenstein's Significance for Developmental Psychology (Berlin/New 202, 353, 356-7 beetle in the box 313
York: Springer, 1992). analytic definition 26, 33, 35, 1 1 3 - 1 4 , BEHAVIOUR AND BEHAVIOURISM 27~8,
120-4, 152 40, 5 5 - 8 , 81, 129, 135, 156-8, 175-7,
Dilman, I., Freud and the Mind (Oxford: Blackwell, 1984). [Features many illu­ ancestral relation 266 185, 288, 313-14, 351
minating comparisons with Wittgenstein.] 'and so on' 149, 265, 328 BELIEF 52, 5 8 - 6 3 , 77, 81, 288-91, 3 5 9 -
Harris, R., Language, Saussure and Wittgenstein (London: Roudedge, 1988). Anscombe, G.E.M. 29, 74, 75 61; see also INTENDING AND MEANING,
Hyman, J. (ed.), Investigating Psychology (London: Roudedge, 1991). [Contains anti-realism 95, 382, 384 THOUGHT/THINKING
ANTHROPOLOGY 3 5 - 6 , 126, 128, 236; see Bentham, J. 86, 318
articles about the implications of Wittgenstein's thought for contemporary
also HUMAN BEING Berkeley, G. 360
psychology.] a priori see analytic/synthetic; philosophy; 'Big Typescript' 23
Kerr, F., Theology after Wittgenstein (Oxford: Blackwell, 1986). [Spells out var­ synthetic a priori BIPOLARITY 17, 21, 6 3 - 6 , 202, 259, 262,
ious implications for theology.] Aquinas, T. 323 331-2, 365
Langer, S. K., Philosophy in a New Key (Cambridge, Mass: Harvard University argument see function bivalence 98-100, 273; see also
Press, 1948). [Applies a theory of symbolism influenced by TLP to pro­ Aristotle 29, 43, 124, 199, 212, 220, 226, BIPOLARITY
241, 292, 294, 300, 318, 340, 354, 362 Blue and Brown Books 23-i, 41, 195, 285
blems in aesthetics.] arithmetic 20, 24, 234 brain 157, 177-9, 181, 242-3, 374
Pitkin, H. F., Wittgenstein and Justice (Berkeley: University of California Press, aspect-blindness 39 Boltzmann, L. 11, 12-13, 77, 220, 341,
1972). [Draws implications for social and political thought.] aspect-dawning 3 6 - 9 343
Tilghman, B. R., But is it Art? (Oxford: Blackwell, 1984). [Discusses whether ASPECT-PERCEPTION 27, 34, 3 6 - 4 0 , 57, Bolzano, B. 43, 171
art can be defined from a Wittgensteinian perspective.] 120, 170; continuous 40 Boole, G. 198, 368
assertion 6 0 - 3 , 3 0 1 - 2 bounds of sense see limits of thought/
Winch, P., The Idea of a Social Science and its Relation to Philosophy (London: assumption 6 1 - 3 language
Roudedge, 1958). [Locus classicus for a Wittgensteinian approach to the ex­ Augustine 25, 41, 242, 277, 285, 295 Bradley, F.H. 59, 116, 189, 316
planation/understanding debate.] AUGUSTINIAN PICTURE OF LANGUAGE 25, Brentano, F. 184, 188
4 1 - 5 , 144, 175, 195, 211, 238, 255-6, Broad, C D . 363
274, 277, 310, 376 broom 115-16, 122, 269

396 397
 INDEX
Brouwer, L. E. J. 21,234-5
Biihler, K. 318
Burge, T. 181
Butler, J. 124
CALCULUS MODEL 23, 67-72, 193-4,
297, 323; see also LOGICAL ANALYSIS
Cantor's diagonal proof 142, 268
Carnap, R. 20, 29, 56, 133, 172, 207-8,
223, 225, 244-5, 310, 335-6, 339, 350
Caroll, L. 134, 216
Cartesianism 28, 45, 156-7, 175-6, 304,
336-7, 348; see also inner/outer
category 189, 214-15, 225, 246-7, 260,
292, 318, 359; see also logical form
CAUSATION 72-6, 171, 189-90, 243;
law/principle of 7 2 - 3 , 342; see also
EXPLANATION; INDUCTION
CERTAINTY 76-81, 171, 173-4, 338
Chomsky, N. 29, 71, 263
Church's theorem 371
cluster 123, 257
Coffey, P. 353
cognitive science 156, 310
Collingwood, R.G. 74
COLOUR 27, 48, 81-4, 276; -exclusion
21, 8 2 - 3 , 106, 356-7; -octahedron
280-1
complex 22, 115-16, 119-20, 256, 269,
271-2
compositionalism 86-8, 231, 260-1; see
also functionalism (in Wittgenstein)
computers 29, 156-9
concept 1 4 , 4 4 , 9 5 , 9 8 , 2 5 8 , 3 1 5 - 1 8 ,
332-3, 368
concept-formation/conceptual change 25,
46-7, 133-4, 135, 155, 228-30, 240
CONSCIOUSNESS 84-6, 349
consistency see CONTRADICTION
constructivism 141-4, 226-31, 2 6 6 - 8
CONTEXTUALISM 32, 86-9, 124, 128-9,
135, 156, 181-4, 256, 260; in Frege 43,
86-7
CONTRADICTION 49, 90-2, 219, 2 3 5 - 6
convention/conventionalism 20, 22, 25,
72, 94, 131-3, 202, 227-8, 343-4; see
also autonomy of grammar
craving for generality 120, 344
CRITERIA 2 6 - 7 , 46, 51, 58, 93-7, 134,
156, 175, 179, 181, 373; defeasibility of
26-7, 9 6 - 7
398
Crusoe, Robinson 309, 329
culture 11, 341, 345; see also AESTHETICS
Davidson, D. 49, 71, 75-6, 89, 128, 178,
390
decision 229, 327
Dedekind, R. 268
Democritus 184
depiction see representation
Descartes, R. 29, 81, 85, 310, 336, 340
description 2 6 , 4 1 - 2 , 4 4 - 5 , 5 0 - 4 , 6 5 ,
109-10, 140, 144-5, 181, 254-7, 275;
vs. avowals 23, 27
desire see INTENnONALITY
determinable 82, 106; see also proposition
system
DETERMINACY OF SENSE 16, 6 8 - 9 , 98-
101, 239, 303
Dewey, J. 379
disposition 57, 183, 325, 328-9, 373,
375-6
dog 361
dogmatism 120, 124, 207, 261-2, 279,
281-2, 297-8
Donnellan, K. 274
doubt 77-81, 307-8, 337-41; makes
sense only within language-game 338—
40
dreaming 340-1
Dummett, M. 71, 197-8, 228, 247, 377
elastic rulers 126
' ELEMENTARY PROPOSITIONS 17,102-7,
116-19, 140-4, 172, 204-7, 237-8,
269-70, 298-302, 366, 368-70; logical
independence of 21, 68, 8 2 - 3 , 102-3,
106, 189, 371
elucidations 249, 255, 275
empiricism 12, 16, 20, 40, 73, 95, 134,
175, 269, 288, 292, 310, 315, 336,
387-90
epistemology see CERTAINTY, INDUCTION,
KNOWLEDGE, SCEPTICISM
Erdmann, B. 198
Ernst, P. 280
essentialism 120-4, 144, 153, 290
ETHICS 18, 28, 107-11, 128, 251-3,
322, 386; and logic/philosophy 12, 107,
330-1; see also SAYING/SHOWING, value
existence see generality
expectation see INTENTIONALITY
EXPLANATION 100, 111-14, 152, 181,
2 4 0 - 1 , 243, 246, 373; aesthetic 33-4;
causal 3 3 - 4 , 37, 72-6, 111, 138, 3 4 2 -
4; come to an end 111-12, 329; genetic
35-6, 111; see also elucidation; ostensive
definition
extensionalism 5 8 - 9 , 140-1, 144
externalism 181-3
FACT 17, 2 1 - 2 , 65, 99, 115-20, 206,
298-301, 303, 315-17; positive vs.
negative 117-19; see also state of affairs
facts of nature see framework
FAMILY RESEMBLANCE 26, 33-5, 65, 73,
100, 114, 120-4, 145, 196, 267, 324,
373-4
falsehood 184-5, 299
Faraday, M. 345
Feigl, H. 20
Ficker, L. von 330, 363
fideism 320-1
finitism 149, 267-8
first-person authority see first/third person
asymmetry
first/third person asymmetry 50-4, 174—
9, 181-4, 288, 290-2, 306-9, 3 5 1 - 2
fly-bottle 2 8 2 - 3 , 350
FORM OF LIFE 26, 32, 124-9, 135-6,
196-7, 322-3, 329, 381
FORM OF REPRESENTATION (Form der
Darsleuung) 22, 46, 129-35, 137, 153,
226, 279, 341, 343-4; alternative 4 8 -
50, 126; in picture theory 129, 214
formal concept 72, 124, 140, 215-16,
246, 266, 293-4, 330-6, 338-9, 341;
see also propositional variable, variable
formal series 141—2, 266
formalism 24, 193, 234-5, 267; see also
nominalism
FRAMEWORK 4 9 - 5 0 , 80, 114, 135-9,
152, 328
Frazer, J. 3 5 - 6 , 278
Frege, G. 11, 13-16, 18-19, 4 3 - 4 , 45,
49, 6 0 - 2 , 6 3 - 4 , 68-9, 8 6 - 7 , 89, 9 0 - 1 ,
98-100, 103, 113-14, 115, 144, 145-7,
164, 175, 193, 198-201, 203-4, 2 0 8 -
11, 212-13, 216, 224, 232, 235, 236-7,
239, 241, 244, 249-50, 254, 257, 258,
262, 264-5, 278, 285-6, 293, 298,
305-6, 315-18, 332-5, 346, 353-4,
399
INDEX
356, 357-8, 361, 363, 365, 368-70,
372-3, 374; -point 60-1
Freud, S. 74-5, 186, 244, 287, 340, 344;
see also psychoanalysis
function 14-15, 98, 103, 208-11, 3 1 6 -
18
functionalism (contemporary) 29, 156
functionalism (in Wittgenstein) 88, 131,
153, 155, 182, 318, 379-81
Gasking, D. 123
game see calculus model; language-game
genealogical tree 282, 2 9 0 - 2
generality 14, 142-3, 145-50, 164-7,
205, 354-5, 369, 371
GENERAL PROPOSITIONAL FORM (allgemeine
Sattform) 17, 21, 26, 102, 120, 140-5,
207, 211, 266, 298-302, 315
genuine duration 38, 63, 179, 289-90,
375
geometry 24, 228, 234
God 109, 179, 229, 251-2, 288, 320-3;
see also trinity
Godel's theorem 92
Goethe, J.W. von 11, 83, 279, 344, 383
Goldbach's conjecture 227, 231
Goodman, N. 302
Goodstein, R.L. 149
GRAMMAR 22, 25, 45-50, 53, 6 7 - 7 2 , 84,
112, 125, 129-30, 132, 135, 150-5,
169, 174, 190, 202, 276, 279-81, 290,
296, 339-40, 378-81, 383-4; depth vs.
surface 69, 154-5, 280; flatness of 2 4 6 -
7; ordinary/school- 18, 23, 71, 150,
153-4, 212, 289, 316; see also FORM OF
REPRESENTATION
grammatical propositions/rules 23, 25,
73, 84, 127, 130-5, 138, 150-5, 186-7,
218-19, 297-8, 324, 336, 367-8; see
also norms of representation
Grelling's paradox 336
Grice, P. 79, 133, 182, 243, 380, 389
harmony between language and reality
22, 186-7, 215, 303, 331
Hart, H.L.A. 100
HamannJ.G. 124
Hegel, G.W.F. 81, 124, 165, 245
Herder, J.G. 124
Heidegger, M. 189, 337
hermeneutics 29, 55, 294, 326
 INDEX
Hertz, H. 11-13, 199, 278, 293, 298,
341, 343, 364
Hilbert, D. 90, 235, 244; see also
metamathematics
Hintikka, J. 198
holism 8 0 - 1 , 89, 133; see also
contextualism
homunculus fallacy 157
HUMAN BEING 29, 5 7 - 8 , 156-9, 162,
362
Humboldt, W. von 87, 124
Hume, D. 11, 63, 7 2 - 4 , 80, 160, 170,
171, 188, 242, 263, 304, 327, 337, 349
Husserl, E. 83, 139, 223
hypothesis 20, 93, 383
I/SELF 65, 160-4, 348-52; as a subject/
object 1 6 1 - 2
Ibsen, H. 375
idealism 175, 310, 315, 337;
transcendental 12, 18, 31, 106, 348-52,
386-90
idea see mental image, mental
phenomena, representation
rDENTrrY 148, 162, 164-9, 232, 328,
369; criteria of 114, 168-9, 312-14;
numerical vs. qualitative 3 0 5 - 6
IMAGINATION 169-70, 262-3
indexicals 43, 102, 254-6, 269
INDUCTION 171-4, 339-40, 342; law of
171, 355; mathematical 2 6 5 - 7
infinity 149, 235, 267-8, 371; axiom of
148, 198, 265, 355
INNER/OUTER 29, 5 5 - 8 , 8 5 - 6 , 174-9,
288-9; see also AVOWALS,
PHILOSOPHICAL PSYCHOLOGY, PRIVACY
INTENDING AND MEANING SOMETHING
63, 129, 179-84, 249-50, 288-90,
380-1
INTENTIONALITY 3 3 - 4 , 5 5 - 6 , 181, 184-
9, 298-9, 303, 387; see also harmony
between language and reality;
isomorphism
internal properties see internal relations;
logical form of names/objects
INTERNAL RELATIONS 17, 21—2, 75~6,
81, 8 3 - 4 , 134, 189-91, 218, 3 0 2 - 3
interpretation (Deutung) 3 8 - 9 , 190-1,
326-9, 358
intuitionism 24, 234-5
400
introspection 26, 51-4, 176, 288-9,
305-6,
inverted spectrum 310, 313
isomorphism 16-17, 87, 186, 213-15,
249, 300-3, 382
James, W. 63, 85, 176, 180, 192, 239,
246, 288, 359, 361, 375, 388
Jean, J. 345
Johnson, W.E. 274
judgement 5 8 - 6 0 , 298—9; see also picture
theory
Kant, I. 12-13, 16, 18, 20, 24, 27, 29,
32, 47, 7 2 - 3 , 102, 113, 132, 160, 169,
175, 199, 232, 233, 247, 258, 263, 269,
280, 282, 288, 292-3, 296-7, 310,
314-15, 3 2 2 - 3 , 326, 335, 337-8, 3 4 1 -
2, 348-9, 352, 353, 357, 360, 367
Kenny, A. 323, 340
Kierkegaard, S. 109, 251, 321, 386
KINAESTHESIS 192, 388
knowledge 51, 76-81, 171, 173-4, 182,
192, 307-9, 336-41
Kohler, W. 3 7 - 8 , 192, 2 4 2 - 3 , 2 8 7 - 8
Kraus, K. 11
Kripke, S. 46, 168, 257, 276
Kronecker, L. 226
Kuhn, T.S. 48, 343
ladder 18, 334-5, 364
language 27, 47, 67-72, 135, 284, 311;
as a city 197; creativity of 87, 298, 301,
372; crisis of 11; critique of 12, 19, 293;
essence of 120-1; guided by norms/
rules 151-4, 193-6, 277, 323-9, 3 7 7 -
80; ideal vs. natural/ordinary 15, 1 8 -
19, 45, 69, 9 8 - 9 , 165, 200, 211-12,
236, 244-5, 261, 296-7; ideal vs. ideal
notation 19, 203, 224, 346, 369;
primary vs. secondary see
phenomenology; sign- 140, 224, 250,
3 5 8 - 9 , 369; of thought 249-50, 3 1 5 -
16, 358-9; see also practice
LANGUAGE-GAME 23, 79, 8 8 - 9 , 120-5,
129, 151, 193-8; vs. calculus model
6 7 - 9 ; crossing of 196-7, 262, 280, 326;
fictional 23, 41, 194-6, 378-9;
framework of 135-9; motley of 196
Laplace, P.S. de 171
laws of thought 13, 16, 216
Lazerowitz, M. 244
learning see teaching
Leibniz, G.W. 45, 68, 229, 353
Leibniz's law 305
liar's paradox 92, 336
Lichtenberg, G.C. 31, 160-1, 304
limits of language/thought 16, 331, 335,
367; see also NONSENSE
linguistics 29, 43, 69
lion 128
Locke, J. 18, 43, 46, 168, 241, 310, 315,
365, 372
LOGIC 13-19, 21, 25, 65, 107, 130, 150,
198-202, 232, 2 5 1 - 3 , 258, 330, 3 5 3 -
7, 363, 368-71, 385; application of 102,
140, 148, 207, 221; Aristotelian/
traditional 14, 198, 316; formal 53, 202,
353
LOGICAL ANALYSIS 43, 68, 70, 82, 98,
167, 203-8, 2 6 9 - 7 4 , 335, 373; and
philosophy 15-16, 18-19, 2 2 - 3 , 203,
206, 208, 293-4; see also LOGIC,
OVERVIEW
logical atomism 16-17,21-2,102-7,
203-8, 269-74, 302 3
logical compulsion/determination/
machine 72, 228-31, 325-6, 328, 375
LOGICAL CONSTANTS 14, 17, 44, 140-4,
145, 199-200,. 208-12, 354, 370-1
LOGICAL FORM 212-16, 299-303, 3 3 0 -
3; of facts/propositions 17, 22, 87, 116,
140, 203-8, 212-14; of language/
reality 16-17, 215; of names/objects
17, 81, 214-16, 260
LOGICAL INFERENCE 2 5 , 4 6 , 1 3 4 , 1 8 9 ,
198, 201, 216-20, 2 3 9 - 4 1 , 331, 356,
371; vs. empirical 201, 217, 219
logical necessity 16-17, 20, 25, 6 4 - 5 , 72,
8 2 - 3 , 130-5, 189-90, 198-202, 2 1 6 -
20, 2 2 3 - 4 , 228-9, 233-4, 239-41,
262-3, 323-4, 342, 353-7
logical positivism 19-20, 25, 49, 65, 110,
131-2, 202, 227, 239, 241, 274-5, 294,
314, 357, 382, 384
logical relations see INTERNAL RELATIONS,
LOGICAL INFERENCE
LOGICAL SPACE 68, 125, 185, 220-3,
273, 302
LOGICAL SYNTAX 1 6 , 1 8 , 2 2 , 1 3 5 , 1 4 0 ,
199-200, 223-5, 259, 273, 323, 331,
334-5; vs. grammar 6 7 - 8 , 150-1, 225
401
INDEX
logicism 13-15, 19, 24, 167, 198, 232,
234, 264-6
Loos, A. 11, 19
lying see pretence
Mach, E. 11,20
magic 36
Malcolm, N. 29, 243, 340
manner of representation (Darstelltmgsweise)
248; see also FORM OF REPRESENTATION
Marr, D. 243
Marx, K. 21, 81, 126, 322
materialism 26, 29, 156-7, 177-9, 288;
eliminative 287; see also mechanism
MATHEMATICAL PROOF 20, 24~5, 90~2,
94-5, 190, 226-31
mathematical propositions 20, 24, 65,
132, 152, 227, 356-7; vs. tautologies
2 3 2 - 3 , 356-7
MATHEMATICS 13, 24, 48, 134, 193,
198-9, 231-6; applied 193, 228, 235;
foundations of 19, 24, 9 0 - 1 , 235-6;
and philosophy 236, 298
Mauthner, F. 11-12, 19, 197, 335, 353
Mead, G.H. 380
MEANING 4 1 - 5 , 6 3 - 4 , 67, 9 4 - 5 , 112,
115, 179, 193-4, 236-9, 256-7, 273,
327, 3 3 0 - 1 , 382-4; causal theory of
379—80; and grammar 150-1; vs.
perlocutionary effect 111, 182, 3 8 0 - 1 ;
speaker's see intending and meaning; see
also AUGUSTINIAN PICTURE OF
LANGUAGE, MEANING-BODY, sense/
meaning distinction, USE
meaning-blindness 3 9 - 4 0
MEANING-BODY (Bedeutungskijrper) 46, 219,
239-41, 277, 376
meaning of life (Sinn des Lebens) 2 5 1 - 3 ,
320
measuring 68, 126, 134, 135-7, 384
mechanism 3 2 5 - 6
Meinong, A. 43, 203
MEMORY 241-3, 311-13; -reaction 184
mentalism 26, 177, 288, 316, 326, 3 5 7 -
9, 361, 372; see also Cartesianism,
psychologism
mental phenomena (acts/events/
processes/states) 179-81, 242, 289-90,
359-61, 374-5, 3 8 8 - 9
mental images 169-70, 180, 2 4 1 - 3 , 250,
288, 359-60, 372-5,
 INDEX
METALOGIC/-MATHEMATICS/-PHILOSOPHY
90-2, 155, 243-7
metaphysics 17, 20, 65, 132, 190, 199-
200, 261-2, 293-6, 3 2 2 - 3 , 335, 384;
of symbolism 18, 22, 4 5 - 6 , 65, 202,
364
METHOD OF PROJECTION 5 9 , 1 7 9 , 1 8 4 - 5 ,
238-9, 247-50, 303, 326, 347-8, 3 5 8 -
9; see also manner of representation,
pictorial relation
Mill, J.S. 20, 43, 190, 233, 254, 257, 292
mind vs. body 58, 158; causal conception
of 75—6, 390; imponderability/
indeterminacy of 97, 129; see also I/SELF,
INNER/OUTER
mind and machine see human being
model theory 219,223
molecular proposition see elementary
proposition
Moore, G.E. 21, 25, 29, 62, 76-8, 109,
115, 138, 189, 206, 250, 297, 315, 338,
356, 357, 362, 365
Moore's paradox 6 2 - 3 , 202
MYSTICISM 12, 18, 31-2, 202, 251-3,
330-1
mythology 129, 280
NAME 14-15, 4 1 - 5 , 164-5, 205, 209,
214-15, 237-8, 254-7, 346-8; proper
42, 254-7; logically proper/simple 17,
4 3 - 4 , 87, 167, 254-6, 269-70; -relation
25-6, 89, 238, 255-6, 270, 312-14,
379; see also AUGUSTINIAN PICTURE OF
LANGUAGE
natural history 138,236
natural laws/necessity 7 2 - 4 , 80, 342-3,
387
naturalism 125-7, 137-8, 174, 337
necessary truths see logical necessity
negation 14, 64, 118, 184-5, 210, 2 3 9 -
41, 299; joint 17, 141-3
net-metaphor 343
NicodJ. 120
Nietzsche, F. 12, 120, 280, 320
nominalism 28, 42, 267
NONSENSE 17, 65, 69, 88, 133, 154,
224-5', 258-64, 293-4, 297-8, 332-3,
335, 355; latent vs. patent 154, 259,
263-4; see also pseudo-proposition;
SAYING/SHOWING
no ownership theory 56
402
norm of description/representation 2 4 - 5 ,
4 8 - 9 , 7 3 - 4 , 130-5, 155, 230, 233-4,
2 6 2 - 3 , 312-13, 323-6, 328, 343-4
NUMBERS 15, 123, 193-4, 232, 264-8
Ogden, C.K. 116, 185
OBJECT (Gegenstana) 17, 22, 84, 103-7,
115-19, 125, 214-15, 220-2, 254-5,
269-74, 302-3, 363-4, 366; arguments
for existence of 270-4; material 115,
175, 269, 271, 338-9, 383-4; number
of 147-8, 221; private 312-14;
properties and relations as 103 5
object of comparison 69, 194-5, 245,
279-81
Occam's razor 238
On Certainty 27, 60, 76, 155
open texture 99
operation 141-3, 209-11, 266-7
ordinary language philosophy 29, 72
OSTENSIVE DEFINITION 25, 46, 84, 114,
136, 152, 255-6, 274-8, 339; private
311 — 14; see also elucidations, sample
Ostwald, W. 363
other minds 95-7, 176-7, 383
OVERVIEW 13, 69, 111, 120, 129, 207,
278-83, 344; vs. logical analysis 2 7 9 -
80
paradigm-case argument 330
patterns in the weave of life (Lebensmuster)
128-9, 156, 183
Pascal, P. 321
Peirce, C.S. 368
perspicuous representation (ubersichtliche
DarsteUung) see overview
personal identity see criteria of identity
phenomenalism see acquaintance, sense-
data
'phenomenology' 23, 8 3 - 4 , 310, 383
PHILOSOPHICAL INVESTIGATIONS 24~7, 28-
9, 41, 284-6 and passim
PHILOSOPHICAL PSYCHOLOGY 22, 55,
124, 179, 286-92; after PI 27, 174,
192; see also INNER/OUTER
Philosophical Grammar see 'Big Typescript'
Philosophical Remarks 23, 6 8 - 9
PHILOSOPHY 16, 18, 2 2 - 3 , 27, 243-7,
292-8; activity rather than doctrine 18,
294, 330, 334; anthropological
perspective on 21, 27; a priori 22, 189—
90, 294-6; critical task of 13, 258-9;
293-4, 296-8, 335; discoveries and
theories in 2 2 - 3 , 68-72, 207-8, 2 7 9 -
80, 297, 344; elenctic rather than
deductive 297-8; and grammar/
language 12, 18-19, 2 2 - 3 , 27, 150-4,
247, 258-9; leaves everything as it is
236, 296; problems of 22, 27, 124,
196-7, 244-5, 282-3, 293-8, 344;
reflective nature of 16, 199, 2 8 0 - 1 ,
292-6; vs. science 18, 27, 111, 138,
292-6, 344-5; systematic 282; see also
dogmatism, LOGIC, METAPHILOSOPHY,
metaphysics, NONSENSE, OVERVIEW,
SAYING/SHOWING, therapy
pictorial form (Form der Abbildung) 213-14,
300
pictorial relation 2 4 7 - 8 , 3 0 1 - 3
PICTURE THEORY 16-17, 22, 50, 61, 87,
99, 116, 184-6, 188, 213-14, 237-8,
247-50, 2 7 0 - 1 , 298-304, 317, 353,
365-6
picture/pictoriality 12-13, 22, 170, 3 0 2 -
4, 341; logical 214, 301-2, 317, 357; -
object 36-8; structure of 213-14, 3 0 0 -
1; see also THOUGHT
Plato 42, 93, 109, 113-14, 184, 189,
241, 270, 285, 318, 360
Platonism 16, 18, 20, 24, 28, 4 3 - 5 , 175,
199, 202, 209, 226, 229, 232-3, 267,
292, 296, 326, 353-4, 361, 372-3
Poincare, H. 265
politics 24, 123
Port Royal Logic 18
Post, E.L. 368
practice 21, 125-7, 197, 327-9
pragmatism 81, 126, 173
pretence 53, 97, 129, 177
Prior, A. 219
PRIVACY 2 6 - 7 , 6 5 - 6 , 176-7, 304-9,
310, 351-2; epistemic 304, 306-9, 372
PRIVATE LANGUAGE ARGUMENT 23, 26,
38, 56-7, 86, 175, 250, 277, 309-15,
329, 352, 372, 383
probability see induction
proof-system 228-9
PROPOSITION (Satz) 16-17, 6 3 - 6 , 140-5,
315-19, 345-7, 357-8; common
sense/hinge (On Certainty) 65, 76-81,
127, 138-9, 178-9, 318, 338-40; as
facts 59, 116, 315-17, 366; genuine
403
INDEX
(Aussagen) 20, 93, 107, 382; as move in a
language-game 89, 90, 263-4, 318-19;
vs. names 44, 209, 316; unasserted 61,
301-2, 315
propositional sign (Satzzeichen) 64, 315—
16, 345-6, 358, 368
proposition system (Satzsystem) 21, 68, 82,
106
propositional variable/function 15, 140—
3, 204, 333-4; see also concept, function
pseudo-problems/propositions 13, 17-18,
20, 60, 65, 165, 199, 224, 232, 293-4,
331, 334-5; see also nonsense
psychoanalysis 111, 297; see also Freud
psychological concepts 287-8
psychologism 11, 16, 26, 198-9, 315,
357, 372-3
Putnam, H. 46, 95, 181
quantifiers see generality; logical constants
Quine, W.V. 29, 81, 89, 95, 128, 133-5,
165-6, 169, 227, 241, 277, 292, 314,
327, 367, 377, 379
radical translation 128-9
rails 326
Ramsey, F.P. 19, 21, 82, 134, 148, 167,
172-3, 224, 233, 266-7, 297, 335, 343,
357, 365-6
range (Sfnelraum) 171, 221
rationalism 18, 73, 155, 343
rationality 12, 27, 124, 126-7, 174, 296,
321
reading 375
realism vs. idealism 12, 28, 350, 352; vs.
nominalism 103-5, 123; see also anti-
realism
reasons 74-6, 158-9,
recognition see MEMORY
referring 41-5, 65, 162-4, 274
Reid, T. 169, 337
RELIGION 18, 251, 320-3; see also God,
trinity
relativism 22, 32, 48-50, 110, 126-7,
281-2, 297
Remarks on the Foundations of Mathematics 24,
284, 324
representation (Vorsteltung) 12, 2 4 2 - 3 , 315,
348, 357, 359-60, 372; linguistic/
symbolic 12, 16-19, 21, 184-5, 198,
270-2, 285, 331, 350, 358; see also
 INDEX
INTENnONALITY, METHOD OF
PROJECTION, THOUGHT
Rhees, R. 29
Richards, J.A. 185, 379
river-bed 155
Rorty, R. 282
rule 49, 9 0 - 1 , 99-100, 132, 312; and its
application 190-1, 325-9; of inference
see logical inference; of language 16, 18,
67-72, 150-5, 193-4, 223-4, 355; vs.
rule-formulation 324; see also GRAMMAR;
LOGICAL SYNTAX
RULE-FOLLOWING 26, 129, 136, 157-9,
168, 190-1, 241, 250, 375, 323-9,
380; vs. acting in accordance with a
rule 7 0 - 2 , 158, 324-5; community-view
of 310-11, 327-9; see also practice
ruler see measuring
rule-scepticism 100, 229, 326-8
RusseU, B . 11, 13-19, 21, 24, 28, 45, 55,
58-9, 6 0 - 1 , 63, 64, 68, 73, 86, 91, 98,
102-4, 106, 115, 118, 134, 145-9, 164,
166-7, 173, 185-9, 189-90, 198-202,
203-5, 208-10, 212-13, 223-5, 2 3 2 - 3 ,
235-6, 242, 244, 251, 254-7, 258,
264-7, 269-70, 285-6, 293, 295, 2 9 8 -
9, 310, 315-17, 322, 330, 333-5, 337,
346, 348-9, 353-7, 357, 363, 365,
369-70, 372, 376, 379, 384, 388;
paradox 14, 9 0 - 1 , 198, 223, 2 3 5 - 6 ,
264, 333-4, 336
Ryle, G. 29, 50, 55, 57, 71, 169, 225,
243, 2 6 0 - 1 , 283, 336, 362, 390
same see identity; rule-following
sample 25, 273-7, 312-13
Sartre, J.P. 169
Saussure, F. de 43, 68
SAYING/SHOWING 17-18, 31—2, 107-9,
139, 206, 232, 2 5 1 - 3 , 259, 262, 320,
330-6
scaffolding 135 see framework
SCEPTICISM 27, 7 6 - 8 1 , 171, 173-4, 191,
336—41; see also other minds
Schlick, M. 2 0 , 2 3 , 1 1 8 , 2 7 5 , 3 1 0 , 3 1 3 ,
343, 350
Schopenhauer, A. 11 - 1 2 , ' 18, 27, 31 - 2 ,
85, 107-8, 160, 199, 234, 250, 251,
293, 335, 348-50, 364, 386-7, 389-90
SCIENCE 12-13, 20, 72-4, 94-5, 111,
123, 322, 341-5
404
Searle.J. 158
seeing 3 8 - 9 , 169-70
Sellars, W. 314
semiotics (syntax/semantics/pragmatics)
263-4, 302
sensations 5 1 - 4 , 128-9, 182, 192, 305-9,
311-14
sense 17, 64, 67, 185, 188, 209-10,
236-9, 248-9, 3 0 0 - 3 , 315-17, 383-4;
autonomy of 237, 271, 2 7 3 - 4 , 303;
secondary 3 9 - 4 0
senseless 17, 90, 259, 3 5 5 - 6
sense/force distinction 4 4 - 5
sense/meaning distinction (Sinn/Bedeutung)
17, 44, 87, 237, 255, 345-7; in Frege
14-15, 4 3 - 4 , 86, 236-7, 254
sense-data 20, 23, 269-70, 310, 383; vs.
minima sensibilia of TLP 105-6, 270
sentence see propositional sign,
proposition
sentence-radical 61
Sheffer, H. 142, 356; stroke 141-2
S i a m e s e twins 3 0 5 - 6
SIGN/SYMBOL 248-50, 3 3 0 - 1 , 333-4,
336, 345-8
simple see complex
situation (Sachhge) 116-17; see also state
of affairs
Skolem, T. 149, 266
Socrates 109, 114, 153
SOLIPSISM 12, 8 4 - 5 , 106, 160, 253, 304,
348-52; methodological 23, 160-1,
304-5, 3 5 0 - 2
some vs. all 150
Sorites paradox 100
soul 5 8 - 9 , 160-4, 250, 304, 321, 3 4 8 -
52, 386; see also I/SELF, SOLIPSISM
speech acts 61, 124, 381
Spengler, O. 11, 22, 120, 124, 234, 279,
344
Spinoza, B . 108, 320
Sraffa, P. 1 1 , 2 1 , 2 4
standard of correctness see norm of
depiction/representation
standard-metre 276
state of affairs (Sachverhalt) 17, 65, 116-
18, 185, 188, 2 2 0 - 2 , 299-303, 316-18;
see also FACT, situation
Stoicism 108
Strawson, P.F. 29, 133, r57, 162, 174,
212, 245, 304, 314, 340, 367
style 23, 28, 284-5, 362-3
superstition 36, 322
sun 88
surveyability 229, 278; see also overview
symptoms 9 3 - 4 , 134
synthetic a priori 20, 65, 83, 132, 199,
202, 233, 263, 357
Tagore, R. 251
Tarski, A. 3 6 5 - 6
TAUTOLOGY 17, 20, 65, 90, 131-2, 164,
165, 167, 169, 2 0 0 - 2 , 209, 216-19,
353-7
teaching 50, 57, 112, 194, 246
theory of content 188-9, 313, 315
theory of descriptions 15, 19, 43, 164-7,
203-5, 254, 256-7
theory of meaning 29, 44, 95, 97, 202,
365-6, 377
theory of symbolism 102, 334, 363, 369;
see also representation (linguistic),
PICTURE THEORY
theory of types 15, 2 0 1 - 2 , 223, 232, 258,
264, 3 3 3 - 4
therapy 111, 297; see also psychoanalysis
THOUGHT/THINKING 16, 55, 58, 156-8,
184-9, 248-50, 315-16, 357-62, 3 7 2 -
3, 387, 389; vs. seeing 38-9; in Frege
14, 175, 237, 315, 357; and language
16, 293-4, 347, 353-62
Tolstoy, L.N. 24,251
tool-analogy 379
GjPTtACTATUS LOGICO-PHILOSOPHICUS 15-19,
23, 26, 28, 285, 330, 335, 362-5, and
passim; linguistic vs. ontological reading
of 205, 270, 3 6 3 - 4
training (Abrichten) see EXPLANAnON;
teaching
transcendental 12, 18, 125, 331
transcendental arguments 338, 340
transition period 20-4, 50, 56, 6 7 - 7 0 ,
108-10, 133, 160-2, 231, 373-4, 380;
see also VERIFICAnONISM
trinity 154, 322~3
TRUTH 2 1 - 2 , 53, 120, 132, 144, 365-8;
see also falsehood
truth-conditions/-grounds/-possibilities
171-2, 217-19, 2 2 0 - 3 , 237-8, 355-7,
369-70
405
INDEX
truth-functional composition 17, 21, 68,
131, 217-18, 355-7, 369-71; see also
analytic/synthetic, LOGICAL INFERENCE
TRUTH-TABLE 65, 171-2, 218, 239-41,
346, 355-7, 368-71; see also language
(ideal vs. ideal notation)
truth-value gap see bivalence
truth-values 63-6, 209; in Frege 14-15,
44, 6 3 - 4 , 208-9, 365
Turing, A. 24, 91, 158; -machine 158
trying 3 8 8 - 9
UNDERSTANDING 26, 71, 87~9, 112-13,
124, 372-6; vs. causal explanation 7 5 -
6; in a flash 374, 376
undogmatic procedure see dogmatism
USE 2 5 - 6 , 112, 139, 238-9, 347, 376-
81; vs. form 32, 111, 131, 154; see also
functionalism (in Wittgenstein)
value 31, 107-8, 320, 331
variable 212, 215-16, 354-5; see also
propositional variable
vagueness see determinacy of sense
VERIFICATIONISM 20, 23, 5 0 - 1 , 53, 56,
93, 162, 197-8, 3 5 0 - 1 , 367, 382-5
visual field 105-6, 160, 349-50
von Wright, G.H. 29, 74
Vygotsky, L. 360
Waismann, F. 20, 60, 91, 99, 132, 223,
231, 265, 275, 2 8 1 - 2 , 382
Weininger, O. 11-12, 107, 348
Weyl, H. 235, 383
Whitehead, A.N. 15
WILL 108, 179, 253, 386-90; cosmic
(Schopenhauer) 12, 29, 386; free 387,
390; vs. intellect 12, 27; subject to 38,
289, 375, 389-90; vs. wishing 387, 389
Winch, P. 75
Wisdom, J. 2 9 , 2 0 5
world (Welt) 17, 21, 115-20, 2 5 2 - 3 ,
349, 386; vs. reality (WMkhkeu) 117,
220
world-picture 124, 127, 129, 179; see also
PROPOSinON (hinge-)
Wundt, W. 288
Zen 253